Air filtration system and method therefor

Abstract

An air filtration system comprises a reverse jet dust collector (250) that comprises an airflow inlet (260) to receive unfiltered air (262), a plurality of filters (252, 254, 256) arranged to filter air input into the airflow inlet and an exhaust output (264) arranged to output filtered exhaust air (266). A pressurized air system is configured to apply pressurized air to clean individual filters of the plurality of filters. A differential pressure sensor (268) is configured to monitor a differential pressure between the airflow inlet and the exhaust output whilst a filter is being cleaned. An analytical software processor (270) is arranged to collate data that includes a differential pressure reading; and a learning processor is arranged to process the collated data and adjust at least one parameter of the pressurized air system in response thereto.

Description

[0001] Attorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0002] - 1 - Title: AIR FILTRATION SYSTEM AND METHOD THEREFOR

[0003] Description

[0004] Technical Field

[0005] The technical field relates to an air filtration system and a method therefor. The technical field is applicable to, but not limited to, an air filtration system and a method for adjusting filtration control parameters to maximise the efficiency with which the system is able to clean filters / filter cartridges within the filtration system.

[0006] Background

[0007] Manufacturing sites around the World that generate pollutants in the environment require air filtration systems to cleanse any air that is being ejected into the atmosphere. The size of these air filtration systems ranges from small units containing a small number of filters / filter cartridges (hereinafter referred to as filters) to large systems being housed in large buildings containing many dozens of filters.

[0008] FIG. 1 illustrates a known air filtration system 100. The known air filtration system 100 includes a filter housing 130 with a dirty gas container area 140, which includes a number of air filters 124, adjacent a clean gas container 108 that has a blowtube 110 and a nozzle 112 and is separated from the filter housing 130 by a barrier 120. The clean gas container 108 is connected to a dust emissions probe 104 that detects how clean the gas 106 is emitted from the air filtration system 100. The clean gas container 108 is connected to a compressed air manifold tank 116 via a solenoid valve 114. FIG. 1 illustrates the air filter at rest 128 and the air filter during a cleaning operation 126.

[0009] It is known that filters 124, 126 acquire pollutants, rendering them ineffective overtime. Air filtration systems, such as known air filtration system 100, that fail to clean the air appropriately to meet air purification standards are subject to severe repercussions and can result in site shutdown, leading to thousands of pounds of loss. Depending on the size of these air filtration systems, conducting maintenance can also be problematic as identifying damaged filter components is time-consuming. Furthermore, air filtration system failure is a critical event that requires immediate action to resolve, thereby placing significant pressure on filtration suppliers to resolve issues promptly.

[0010] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0011] - 2 -

[0012] Filter cleaning systems operate in environments where pollutants are created during the operational process of a business. These pollutants are vast and varied, including elements such as dust particles 138, following the production of pharmaceutical products, fumes from welding operations, paint dust and fumes from spray booths, fragments from milling and cuttings sites, such as wood and metal, etc. These pollutants, which may be dry, wet, greasy, fine or large, are removed from the immediate working vicinity through air extraction systems that ultimately eject this air into the atmosphere. Government legislation mandates that these pollutants be filtered out of the air before being expelled into the atmosphere. To this end, air filtration systems are mandated to be incorporated as part of the air cleaning process.

[0013] As with any filtration system, the pollutants that are removed from the air are done so using various types of materials designed to remove particular types of particles. The particles adhere to the membrane of the filter 124, causing them to clog up and hinder the air from flowing through the air filters. Differential pressure sensors 102 across the filter compartment determine the degree to which these air filters have been clogged. In essence, air 134 being pulled through a dirty filter system via a dirty gas inlet and a diffuser 132 and a filter 124 will produce higher pressure air 118 on the inlet side of the air flow than on the outlet side. Many air filtration systems set thresholds and use this differential pressure reading to trigger a mechanism to clean the filters, typically through a timer / control panel 122 connected to the compressed air manifold tank 116, when the pressure across the whole system exceeds the threshold value. Alternatively, a cleaning event may be triggered by simply implementing a cyclic cleaning repetition that runs irrespective of the differential pressure across the filter system.

[0014] The mechanism that is employed to clean the filters, which are sometimes typically cylindrical filters, utilises short bursts of high-speed compressed air, which is blasted into the centre tube of the filter at high pressure, effectively causing a shock wave on the filter that distorts it, thereby causing the pollutant particles to fall off the air filter. Given the highspeed energy of the compressed air, the dirty particles are forced against the environmental air flow and pushed into a dust collection hopper under the filter unit. This effectively cleans the filter and allows the normal volume flow of environmental air through the system.

[0015] The high-pressure air is built up as a volume of compressed air in a manifold that is pressurised by a stand-alone compressor fitted to (but not considered to be a part of) the

[0016] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0017] - 3 -air filtration system, often not necessarily placed near the air filtration system. The manifold is the vessel that holds the compressed air, so is generally considered a part of the filter cleaning system as the compressed air is used to clean the filters. The compressor is required to ‘charge’ the manifold and when the manifold is at peak charge, the compressor is switched ‘off’. When the manifold pressure drops, the compressor must switch back ‘on’. Solenoid valves 114 activate the bursts of high-energy air into the filter, controlled by a timer controller. Whenever a valve is activated, the pressure within the manifold relaxes as the air is ejected into the filter. Therefore, the cleaning effectiveness of the blast is limited by the remaining pressure of the manifold. As such, the valves are timed to remain open for short intervals (usually within the range of 100-250msec) after which the air pressure is no longer viable as a cleaning agent. Thereafter, the manifold must re-compress before it can become effective again.

[0018] The size of these cleaning systems ranges from units that may contain a single filter to significantly larger filter systems potentially containing over one hundred filters. The number of filters is determined by the volume of air required to be removed from the working environment, or by the types of pollutants that need to be filtered before being ejected into the atmosphere. All these filters need to be cleaned using the mechanism described above.

[0019] In essence, the inventors have identified operational challenges with the above. A first operational challenge is that the current controller systems manage the cleaning cycle in one of two ways: a first way is that they monitor the differential pressure (using differential pressure sensor 102) between the dirty and clean sides of the filter housing 130 and trigger a cleaning event when the pressure differential exceeds a particular value. A second way is that the controller is set to continuously run through a cleaning cycle. The cleaning cycle itself is managed primarily through two parameters. A first parameter is the duration that the solenoid valve 114 is to remain open is the active time that a filter element is being cleaned. A second parameter is the duration required before the next solenoid valve 114 can be activated to allow the compressed air building up in the manifold to reach a suitable level.

[0020] Once these two parameters are set, they remain fixed, simply because it is not cost-effective for suppliers of these air filtration systems to continuously monitor and modify these parameters on a permanent basis. Many of these sites may be located a considerable distance from the system supplier. Further to this, the controller units form a small percentage of the revenue received from these systems, and the motivation to upgrade them to support modern techniques and technologies has not yet been enough within the

[0021] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0022] - 4 -industry to warrant a significant desire to do so. This has left suppliers of these air filtration systems with very basic tools to modify any controller parameters, and, to date, requires the physical presence of a technician onsite to do so.

[0023] The inventors have further recognised and appreciated that issues with the current solution can also arise because of the diverse nature of the industry. Pollutants that these air filtration systems need to address do not lend themselves to a ‘one-size-fits-all’ approach, yet there is no database to which a supplier can refer to that provides up-to-date, accurate information on the best parameters for any given filter design and particle type. Today’s best practices are thus based on experience, rather than up-to-date performance management systems and processes. Furthermore, the parameters of the air filtration system change as the performance of the filters 124 alters due to dirt build-up that isn’t removed (e.g. greasy particles may build up quicker than dry particles), as well as the effects of wear and tear on the filters 124. Eventually, this can lead to pollutants not being removed from the air, thereby finding their way into the atmosphere. In such cases, the site may be required to shut down until the filter(s) 124 can be replaced. To mitigate this eventuality, suppliers set up maintenance contracts based on servicing the units at fixed intervals. These services are not necessarily the most effective or efficient use of time or resources, resulting in potentially unnecessary costs to both the supplier and end-user.

[0024] The inventors have further identified that many types of filter materials, designs and overall filter systems are available based on the varying types of pollutants and volume of air that is required to be cleaned. They have recognised and appreciated that one of the key issues with current designs is that, although the entire system's performance can be inferred using the differential pressure sensor across the filter system, each filter’s 124 individual effectiveness and contributions cannot be determined. Therefore, the failure of any single component leads to the failure of the entire air filtration system 100. A need therefore exists for an air filtration system and a method for adjusting filtration control parameters to maximise the efficiency with which the air filtration system is able to clean such filters.

[0025] Summary:

[0026] In a first aspect, an air filtration system is described that comprises: a reverse jet dust collector that comprises an airflow inlet to receive unfiltered air, a plurality of filter arranged to filter air input into the airflow inlet and an exhaust output arranged to output filtered exhaust air. A pressurized air system is configured to apply pressurized air to clean

[0027] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0028] - 5 -individual filters of the plurality of filters. A differential pressure sensor configured to monitor a differential pressure between the airflow inlet and the exhaust output across a reverse jet dust collector whilst an individual filter is being cleaned. An analytical software processor operably coupled to the differential pressure sensor and arranged to collate data that includes a differential pressure reading from the differential pressure sensor. A learning processor is operably coupled to the analytical software processor and arranged to process the collated data and adjust at least one parameter of the pressurized air system in response to the processed collated data. In this manner, an air filtration system (and a method) for adjusting filtration control parameters is described to maximise the efficiency with which the air filtration system is able to clean such filters.

[0029] In some optional examples, the learning processor may be further arranged to perform at least one of the following: predict at least one failure event of the pressurized air system; adjust a cleaning cycle of the plurality of filters of the pressurized air system; adapt at least one component in the pressurized air system to affect an efficiency parameter of the air filtration system. In some optional examples, the learning processor may be further arranged to dynamically adjust at least one of the following: a solenoid activation time, a manifold recovery time.

[0030] In a second aspect, a method for adjusting filtration control parameters of an air filtration system is described. The method comprises: receiving unfiltered air, by a reverse jet dust collector that comprises an airflow inlet and a plurality of filters; applying pressurized air to clean individual filters of the plurality of filters; outputting filtered exhaust air by an exhaust output of the reverse jet dust collector; monitoring a differential pressure between the airflow inlet and the exhaust output across the reverse jet dust collector by a differential pressure sensor whilst an individual filter is being cleaned; collating data that includes a differential pressure reading from the differential pressure sensor; processing the collated data by a learning processor; and adjusting at least one parameter of the pressurized air system in response to the processed collated data.

[0031] In a further aspect, an air filtration system is described that comprises: a reverse jet dust collector that comprises an airflow inlet to receive unfiltered air, a plurality of filter arranged to filter air input into the airflow inlet and an exhaust output arranged to output filtered exhaust air. A pressurized air system is configured to apply pressurized air to clean individual filters of the plurality of filters. A differential pressure sensor configured to monitor a differential pressure between the airflow inlet and the exhaust output across a reverse jet

[0032] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0033] - 6 -dust collector whilst an individual filter is being cleaned. An analytical software processor operably coupled to the differential pressure sensor and arranged to process received continuous air filtration system behaviour data that includes a differential pressure reading from the differential pressure sensor; and dynamically adjust in a real-time manner at least one filter cleaning parameter in response to the processed data. In this manner, an air filtration system (and a method) for adjusting filtration control parameters in a real-time manner is described to facilitate dynamically re-configuring the parameters of the air filtration system in order to maintain the most effective and efficient air-cleaning capabilities, to counteract the filters degrading over time, thus extending and optimising the air filtration system's overall performance.

[0034] The concepts described herein provide an air filtration system and a method for adjusting filtration control parameters to maximise the efficiency with which the air filtration system is able to clean filters, as described in the accompanying claims. Specific example embodiments are set forth in the dependent claims. These and other aspects will be apparent from, and elucidated with reference to, the examples described hereinafter.

[0035] Brief description of the drawings

[0036] Further details, aspects and example embodiments will be described, by way of example only, with reference to the drawings. In the drawings, like reference numbers are used to identify like or functionally similar elements. Elements in the figures are illustrated for simplicity and clarity and have not necessarily been drawn to scale.

[0037] FIG. 1 illustrates a known air filtration system.

[0038] FIG. 2 illustrates one example of a stand-alone air filtration system with valve activity detection for example to provide timing as a basis for aligning filter measurements in a time domain, adapted in accordance with some examples.

[0039] FIG. 3 illustrates one example of an integrated air filtration system where a monitoring unit detects activity on each solenoid valve to determine a pressure profile across each filter before, during and after a cleaning event, in accordance with some examples.

[0040] FIG. 4 illustrates a timing example of how differential pressure is collated, in accordance with some examples.

[0041] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0042] - 7 -

[0043] FIG. 5 illustrates examples of air filtration system architectures connected to a cloud-based server, for example including a neural network to assess the data obtained from the implementation of any of the other figures, in accordance with some examples.

[0044] FIG. 6 illustrates an example flowchart to determine an efficient system setup that maximises the filters' cleaning, in accordance with some examples.

[0045] FIG. 7 illustrates an example flowchart of a machine learning process of a neural network to determine an efficient system setup that maximises the filters' cleaning, in accordance with some examples.

[0046] FIG. 8 illustrates a yet further example of a stand-alone air filtration system that captures manifold pressure to determine a pressure profile across each filter before, during and after a cleaning event, in accordance with some examples.

[0047] FIG. 9 illustrates a yet further alternative example of an integrated air filtration system that captures manifold pressure to determine a pressure profile across each filter before, during and after a cleaning event, in accordance with some examples.

[0048] FIG. 10 illustrates a timing example of how manifold pressure is collated to determine a pressure profile of FIG. 8 or FIG. 9 in utilising compressed air for cleaning filters, in accordance with some examples.

[0049] FIG. 11 illustrates a yet further example of an air filtration system connected to a neural network with an automated feedback mechanism to rapidly adjust a cleaning operation of each filter, in accordance with some examples.

[0050] FIG. 12 illustrates one example of an operation of a neural network adapted to support the examples described herein.

[0051] FIG. 13 illustrates one example flowchart for adjusting filtration control parameters of an air filtration system and / or adjusting in a real-time manner at least one filter cleaning parameter.

[0052] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0053] - 8 - Detailed description

[0054] The current arrangement to assess a performance and / or clean an air filtration system requires on-site intervention and analysis in order to determine a performance of the cleaning system, as maintenance or fault analysis requires a person (e.g., an air filtration engineer) to work through the system, element by element, on site. In particular, both the pressure of the compressed air in the manifold as well as the pressure differential across the filter system can only be determined across the entire filter system. These are two distinct features of the system, i.e., the differential pressure increases as the air through the filter becomes impeded by dirty filters, and the compressed air is required to be sufficiently charged in order to clean said filter. Thereafter, a clean of the totality of filters is required to be undertaken in a time-consuming and somewhat disruptive manner. In contrast to the current approach examples herein described provide an air filtration system that is able to monitor and analyse a behaviour of individual filters and determine the effect on the associated filter cleaning system’s respective performance as well as monitoring how efficiently the cleaning mechanism is cleaning individual filters. Some examples described herein propose a use of Internet of Things (loT) devices that are configured to monitor and analyse the performance of each individual element. In some examples, this may include monitoring the differential pressure variance caused at the moment when each filter is cleaned. This approach determines the effects by detecting the activity of individual solenoid valve activity and / or inactivity and the effect on pressure in order to determine cleaning event behaviour over time accurately. In this manner, a profile of the effectiveness of each filter and cleaning component may be developed as each filter progresses from being adequate to ineffective as the filter(s) degrade(s). In some examples, by tracking the differential pressure as a function of the solenoid activation time and manifold pressure, the effectiveness of the cleaning mechanism may be advantageously determined on a per-valve / filter set basis.

[0055] In some examples, the receiving device, e.g. a cloud-based tool, may include a learning processor where artificial intelligence techniques may be used to analyse the effects of various parameters on the performance of the air filtration system. In some examples, the learning processor may have been trained with information about the types of pollutants versus a size and design of various air filtration systems in order to determine how best to alter the parameters of any given air filtration / cleaning system. These adjustments may then be used to further improve any machine learning algorithm performed by the learning processor when adjusting the cleaning process parameters or identifying alternate cleaning

[0056] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0057] - 9 -techniquesto improve effectiveness and efficiency. In some examples, it is envisaged that such adjustments may include altering a duration of each cleaning air blast from the solenoids, a sequence order applied to the solenoids, and / or a timing of the adjustment based on learnt filter differential-pressure and manifold pressure profiles.

[0058] In some examples, it is envisaged that valve / solenoid diameter may be used as part of a calculation for analysing cleaning performance. The inventors have recognised that valve / solenoid diameter impacts the following:

[0059] (i) The level of energy that is brought to bear on the filter, i.e., smaller valve diameters will reduce the amount of energy that can be used to clean the filter.

[0060] (ii) A bigger valve diameter will cause the manifold pressure to drop more rapidly. This, therefore, affects the useful time available for cleaning the filter. Once the manifold pressure drops below a certain threshold, there is no longer effective cleaning taking place. Hence, keeping the valve open at this point is wasteful.

[0061] (iii) Better understanding of the performance of various valve diameters in particular designs aids more effective designs of future cleaning / air filtration systems. This includes: (i) developing better insights into what are the most appropriate valve diameters based on a given design, (ii) what is the most effective volume of air required in the manifold for a specific compressed pressure, and (iii) how best to operate the valves.

[0062] In some examples, pressure monitoring may be achieved using a differential pressure sensor (such as differential pressure sensor 268 in FIG. 2. However, in other envisaged examples, it is anticipated that pressure (or differential pressure) monitoring may be achieved using other sensors, piezoresistive sensors, capacitive pressure sensors, airflow sensors, or other suitable devices, as would be appreciated by a skilled artisan.

[0063] Some examples of the air filtration system introduce one or several features designed to optimise filter cleaning efficiency through real-time monitoring and adaptive control mechanisms, including one or more of:

[0064] (i) Monitoring the activation of each individual solenoid valve in order to track cleaning events accurately in real-time.

[0065] (ii) Capturing and analysing differential pressure profiles associated with each cleaning event to assess an individual filter’s performance.

[0066] (iii) Transmitting operational data to, say, a cloud-based server for remote processing and analysis.

[0067] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0068] - 10 - (iv) Applying machine learning algorithms to the collected data in order to predict future failure events, optimise cleaning cycles, and enhance system efficiency.

[0069] (v) Dynamically adjusting cleaning parameters such as solenoid sequencing, activation timing, and pneumatic blast energy based on continuous analysis of system behaviour.

[0070] (vi) Establishing a communication link using technologies such as Wi-Fi™, cellular, Bluetooth™, or Radio to facilitate seamless integration with the cloud infrastructure.

[0071] In this manner, examples herein-described provide a filtration system that evolves to maintain peak performance, reduce maintenance costs, extend filter life, and ensure compliance with environmental regulations.

[0072] In this manner, examples described herein may improve air filtration system efficiency in one or more of the following ways:

[0073] (i) Through visualisations and early warning systems, for example based on parameters input by the manufacturer or installer. It is envisaged that these may be monitored remotely and / or provide notifications when thresholds are met;

[0074] (ii) Improved operationalisation may be achieved using machine learning to predict performance based on historic information. It is envisaged that this may also be used to recommend a best air filtration system setup for a particular type of environment; and (iii) Through learning mechanisms, for example whereby the system parameters may be manipulated through artificial intelligence (Al) algorithms and the resultant performance improvement or degradation may be analysed in order to provide insight into best practice for the specific environment.

[0075] Examples herein described, it is envisaged that, using the additional information obtained to improve overall cleaning efficiency and design, may also prevent the compressor being switched ‘on’ unnecessarily, thereby avoiding wasting energy.

[0076] Throughout this document and the figures, the term ‘reverse jet dust collector’ is used to encompass at least the complete unit to be cleaned. Furthermore, throughout this document and the figures, the term ‘filter’ is used to encompass the filter and various envisioned filter types as well as other forms of ‘dust collector’. A list of ‘filter’ types that are envisioned as benefitting from the concepts described herein, in a reverse jet dust collector, include:

[0077] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0078] - 11 - 1. Cartridge Filters: typically cylindrical filters that are efficient for fine dust, and easy to clean.

[0079] 2. Bag Filters: typically long, cylindrical bags, great for handling heavy dust loads. 3. Pleated Filters: Similar to cartridge filters, but with a larger surface area for dust capture.

[0080] In addition to the above main filter types, some envisioned specialty filter applications (within the described reverse jet dust collector) include:

[0081] 1. Envelope Filters: typically flat filters often used in compact spaces.

[0082] 2. Ceramic Filters: typically used for high-temperature applications.

[0083] 3. High-efficiency particulate air (HEPA) filters: typically used for capturing very fine particles, though less common in industrial dust collection.

[0084] Referring now to FIG. 2, illustrates one simplified example of a stand-alone air filtration system 200 with valve activity detection, for example to provide timing as a basis for aligning filter measurements in a time domain, in accordance with some examples. The simplified example of a stand-alone air filtration system 200 of FIG. 2 is focused on the control mechanisms for cleaning the filters, rather than the operations of the filter system itself. A skilled artisan will recognise that the standard features of FIG. 1 are also contained, but not shown in FIG. 2 to avoid obfuscating the details of the invention, such features including a clean gas container 108 that has a blowtube 110 and a nozzle 112, a barrier 120, a dust emissions probe 104 that detects how clean the emitted gas 106 is, a compressed air manifold tank 116, etc.

[0085] The air filtration system 200 includes an air filtration cleaning system 201 that includes a solenoid supply 210, which can be a DC or AC voltage, that controls the operation of (in this simplified example) three solenoid enable switches 212, 214, 216, responsive to respective solenoid enable signals 202, 204, 206. The solenoid switches 212, 214, 216 are connected to respective solenoid coils 242, 244, 246 and energizing valves 282, 284, 286 respectively. Activity monitoring points 222, 224 and activity detector 226 are illustrated where current is flowing through the coil is measured, and the voltage presented by the controller 320 is detected by a controller, for example in a form of an analytical software processor 270. A solenoid activation mechanism that uses coils 242, 244, 246 to dictate whether (or not) a specific valve is activated and open to receive and expel short bursts of compressed air from a pneumatic pressure tube 232, 234, 236 connected to a manifold (not shown). When the coils 242, 244, 246 are energised it moves the respective valve (282, 284 and 286) to the ‘open’ position, thus dictating whether (or not) a specific valve is activated. It is

[0086] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0087] - 12 -noteworthy that the differential pressure through the filter system is permanently monitored so time-based metrics of the rate with which these filters are being clogged up can be measured. It is also noteworthy that the knowledge of how each individual filter is behaving can only be determined at the point at which it has been cleaned, i.e., when the solenoid is activated.

[0088] The air filtration system 200 includes a reverse jet dust collector 250, which is a contained area of unfiltered air and three filters, with a dirty gas container area that includes a number of air filters 252, 254, 256 (with three shown for simplicity purposes only). The reverse jet dust collector 250 includes an unfiltered airflow inlet 260 to receive unfiltered air 262 and an exhaust output 264 arranged to output filtered exhaust air 266.

[0089] In FIG. 2, a first air filter 252 is shown as being cleaned with short bursts of high-speed compressed air, which is blasted into the centre tube of the filter 252 at high pressure, by a first solenoid actuator 282. The short bursts of high-speed compressed air effectively cause a shock wave on the filter 252 that distorts it, thereby causing the pollutant particles to fall off the air filter 252 membrane. In this manner, the filter 252 is cleaned, thereby removing any pollutants / particles that have adhered to the membrane of the filter 252. This effectively cleans the filter 252 and allows the normal volume flow of environmental air through the air filtration system 200.

[0090] The second up to an nth filter 254, 256 are shown at rest in this phase during a cleaning operation. Unfiltered air 262 is pulled through a dirty filter system contained in reverse jet dust collector 250 via the unfiltered airflow inlet 260 (and a diffuser (not shown)). The filter 252 produces higher pressure air on the inlet side of the air flow than on the exhaust output 264. As air passes through the filtration unit / reverse jet dust collector 250, a pressure difference builds up across all the filters 252, 254, 256 as a function of a level of pollutant that has collected in the filters 252, 254, 256, in that higher pressure air is monitored on the airflow inlet 260 than on the exhaust output 264. A differential pressure sensor 268 is configured to monitor the differential pressure between the airflow inlet 260 and the exhaust output 264 across the reverse jet dust collector whilst the filter 252 is being cleaned and provide this differential pressure reading to an analytical software processor 270.

[0091] In examples described herein, a use of the information using ML and Al may allow a determination of the most appropriate cleaning method to be used when the system is triggered based on developing an understanding of the most efficient use of the various

[0092] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0093] - 13 -parameters applied during the cleaning cycle. It is envisaged that these parameters may include:

[0094] (i) How long should the valve be kept open (also determined by the manifold pressure performance);

[0095] (ii) How many times should the activation of each valve be repeated to maximise cleaning performance, also taking into account the associated degradation that this may cause on the filter itself.

[0096] (iii) How long to wait before the manifold pressure is high enough to activate a valve. Here, it is envisaged that activating it too quickly may reduce the impact of the blast of air used to clean the filter, whereas taking too long may result in wasted energy being used by the compressor.

[0097] (iv) Determination of what would be the most appropriate differential pressure points to start a cleaning cycle. In existing systems, this is a fixed value that is inserted at installation. However, there is no understanding of whether this is the best point at which to trigger a cleaning event, or whether a more dynamic approach to cleaning would produce better results. Thus, employing the concepts described herein may identify a better threshold to trigger a cleaning event, which may change based on the type of material used, the length of time the filters have been in operation, the type of pollutant in the air, etc.

[0098] In accordance with examples described herein, the analytical software processor 270 also includes valve activity detection / monitoring and therefore may receive inputs from the respective valves identifying those valves that are open. Activity monitoring points 222, 224 and activity detector 226 are illustrated where current is flowing through the coil is measured, and a voltage is present at the controller / analytical software processor 270. The valve activity detection provides timing as a basis for aligning respective filter measurements in the time domain, rather than the current known system of FIG. 1 that is solely focused on monitoring a pressure differential across the totality of the filter of the filtration unit.

[0099] In some examples the communication may be performed using, say, modern wide-area networking techniques 520, such as: cellular, Internet of Things (IOT), Bluetooth™, Wi-Fi™ or any other such technology to connect to, say, cloud-based tools employing cloud-based analytical software (as described with reference to and in FIG. 5 and FIG. 6). In some examples, it is envisaged that a cloud-based tool employing a cloud-based analytical software processor may be arranged to change the parameters of say a controller / timer,

[0100] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0101] - 14 -using a communications interface (for example as illustrated in FIG. 3). Thus, Al can still be applied through the simple understanding that can be developed through understanding the behaviour of a number (or indeed a large number) of air filtration system components and parameters, which can form the basis for setup and design of a system.

[0102] Referring now to FIG. 3, one example of an integrated air filtration system 300 is illustrated where a monitoring unit detects activity on each solenoid valve in order to determine a pressure profile across each filter, i.e., before, during and after a cleaning event, in accordance with some examples. The example of an integrated air filtration system 300 is similar in many respects to the above-described stand-alone air filtration system 200 of FIG.

[0103] 2, albeit in FIG. 3 there is a direct communication between the existing controller processing unit, and intelligent cleaning algorithms and therefore supports a direct integration with a supplier’s controller. Therefore, like functions and functionality between FIG. 2 and FIG. 3 will not be repeated here for the sake of clarity and to not obfuscate the explanations to a reader.

[0104] In the air filtration system 300, the communication controller 320 is modified to communicate the various parameters of the air filtration system 300 to an analytical software processor 270. In some examples, the analytical software processor 270 may be configured to provide predictions about an expected window for servicing the filters or other air filtration system components so that suppliers and end-users are better able to manage their operations effectively. In some examples, the analytical software processor 270 may be configured to enable early warning of potential issues that may lead to the air filtration system 300 becoming ineffective, which would ordinarily lead to costly shutdowns as well as potentially illegal pollutant egress into the atmosphere. In some examples, the analytical software processor 270 may be configured to provide a mechanism by which advanced air filtration systems may dynamically re-configure the parameters of the air filtration system 300 in order to maintain the most effective and efficient air-cleaning capabilities, as the filters degrade over time, thus extending the air filtration system's overall performance. In some examples, the analytical software processor 270 may be configured to provide an ability to feed back into the communication controller 320, such that energy efficiency may be improved since control elements, such as the compressor and valve solenoid durations may be managed based on real-time fed back information of the air filtration system rather than default parameters. Since both these elements require power, unnecessary usage results in excessive energy bills for the end user. It is envisaged that examples described herein may also reduce the carbon footprint for operating such an air filtration system, as there is

[0105] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0106] - 15 -no longer a need for unnecessary travel, unnecessary plant operations or outages, reduced likelihood of pollutants finding their way into the environment etc.

[0107] In some examples, a direct communication link 269 connects sensors to a monitoring system and provide component states and values, as well as a monitored differential pressure sensor level between the pressure sensor 268 and communication controller 320. In a localised implementation, the direct communication link 269 may use a communication medium such as UART serial communications 0.5v - 4.5V analogue, 4-20mA analogue, I2C digital, USB, RS232, RS485, Ethernet. In some examples, it is envisaged that sensors may be connected to the internet and eventually the cloud using a wireless technology such as Bluetooth (BT™) or Bluetooth Low Energy (BLE™), Zigbee™ IOT-NB, LTE-M, LTE 4G or other forms of digital wireless communications or medium. In some examples, it is envisaged that sensors may be connected to the internet and eventually the cloud, via the controller running intelligent cleaning algorithms.

[0108] Referring now to FIG. 4, a timing diagram 400 example of how differential pressure is collated is illustrated, in accordance with some examples. In these examples, a skilled artisan will recognise that the terms A1 to An actually refer to the measurement of both voltage and current through the solenoid. In order to determine if the valve is indeed open, the current is monitored through the coil. However, it is also understood that the valve should be ‘open’ if there is a voltage present on it, and this is used to aid fault finding. If there is indeed current passing through the solenoid, then it is assumed that the valve has been activated. The timing diagram 400 includes a first waveform 405 illustrating a proxy measurement ‘AT that indicates that a first valve should be active for a first period of time S1 410 that is ‘high’ when a short blast of compressed air is expelled into a first filter. The timing diagram 400 includes a second waveform 415 illustrating a proxy measurement ‘A2’ that indicates that a second valve should be active for a second period of time S2420 that is ‘high’ when a short blast of compressed air is expelled into a second filter. The timing diagram 400 includes a third waveform 425 illustrating a proxy measurement ‘An’ that indicates that an nth valve, e.g., a third valve, should be active for a third period of time Sn 430 that is ‘high’ when a short blast of compressed air is expelled into the nth filter. It is envisaged that a measurement of current may also provide a mechanism for remote fault finding, for example if there is a failure to clean a particular filter. It is envisaged in other examples, a different approach (other than proxy current measurements) may be employed to determine whether (or not) a valve is ‘active’.

[0109] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0110] - 16 -

[0111] A fourth waveform 435 illustrates the consequent operation of the differential pressure measured by the differential pressure sensor 268. As illustrated, the fourth waveform 435 includes periods 440, 445, 450 of reducing differential pressure that coincide with one of the solenoid valves being sequentially opened to short blasts of compressed air input to a respective filter. In this manner, by monitoring the differential pressure and the valve opening periods, the controller or analytical software processor is able to determine and collate the filter differential pressure at the beginning of each valve’s cleaning event, the filter differential pressure at the end of each valve’s cleaning event, using the duration of the event based on, and obtained using, the valve activity signals.

[0112] In some examples, the differential pressure profiles and associated solenoid activation durations may be collated with other system information, such as time, date, filter activity, intensity, etc., and securely forwarded to an analytical software processor. In some examples, the analytical software processor may be located in one or more cloud-based servers and process data related to each filter activation event. In some examples, the analytical software processor may be a learning processor with a neural network and arranged to use machine language to predict air filtration system future performance and operational details. In some examples, as the volume of information expands, the information is used with various machine learning tools and methods in order to determine performance and operational trends for each filter. In some examples, the analytical software processor may form a basis of a scalable analytics platform. In some examples, by leveraging machine learning algorithms, the cloud-based server(s) process(es) this information to generate suggestions for optimising the operational parameters of each filter of each filtration system, thereby enhancing overall efficiency and performance across diverse industrial environments.

[0113] Referring now to FIG. 5, examples of air filtration system architectures 500, 550 connected to a cloud-based server, for example including a neural network to assess the data obtained from the implementation of any of the other figures, are illustrated in accordance with some examples. In accordance with examples described herein, the air filtration system architectures 500, 550 include the analytical software processor 270 of FIG. 2 or FIG. 3 noting that the valve activity detection and receiving inputs from the respective valves identifying those valves that are open is not shown to avoid obfuscating the concepts described with regard to a cloud-based server implementation. Again, activity monitoring

[0114] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0115] - 17 -points 222, 224 and activity detector 226 are illustrated where current is flowing through the coil is measured, and a voltage is present at the filter / analytical software processor 270. In these examples the analytical software processor 270 includes logic or a processor 515 configured to perform measurement and reporting of air filtration system parameters.

[0116] A first air filtration system architecture 500 illustrates the measurement and reporting of air filtration system parameters using any suitable communication technology / interface to a cloud-based server 530 that includes a cloud storage database 532. In this example, the cloud storage database 532 is configured to store one or more of the following: a filter pressure profile, a manifold pressure profile, valve activity, pollutant classification data, filter design parameters, etc. In some examples, the cloud-based server 530 may be made available to operators and management staff of the air filtration systems and configured to provide the latest suggested parameters to be setup for any filter system that includes analytical software processor 270 based on the pollutant types and filter design parameters.

[0117] In some examples, the cloud-based server 530 may be configured with neural network capability, as described with reference to FIG. 12, where data analytics 534 may be extracted from cloud storage database 532 and provided to a learning processor 538 that uses the neural network to provide air filtration recommendations to analytical software processor 270. In accordance with some examples, the cloud-based server 530 with the neural network is configured to learn from the data provided to it and continuously determine the most efficient system setup that maximises the filters' cleaning operations.

[0118] A second air filtration system architecture 550 illustrates the information 570 passed from the cloud-based server 530 to an accessible on-line, repository 560. The information 570 may include parametric suggestions for the analytical software processor 270 to apply, including one or more of the following: valve duration, reducing or extending the electronic opening pulse time of each solenoid valve, manifold recovery time, which is the time taken for the compressed air pressure to increase to the set air pressure, filter pressure thresholds, etc. In some examples, it is envisaged that such parametric suggestions may be useful to improve future air filtration system designs.

[0119] Referring now to FIG. 6, illustrates an example data flow 600 is illustrated to determine an efficient system setup that maximises the filters' cleaning, in accordance with some examples. The example data flow 600 includes a plurality of metrics 610, such as filter

[0120] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0121] - 18 -metrics 612-618 that are provided to a database 626. The database 626 communicates performance data to / from a learning processor 624, which in some examples employs machine learning to predict cleaning system parameters and an associated performance. The learning processor 624 also communicates to / from a database 622 that includes information that categorizes pollutant type and filter design. The pollutant type and filter designs are provided by vendors of pollutants and / or filters / filter systems, 632-636.

[0122] Referring now to FIG. 7, an example flowchart 700 of a machine learning process of a neural network to determine an efficient system setup that maximises the filters' cleaning is illustrated, in accordance with some examples. The flowchart starts at 705, with retrieving, for example by a database performing a query search (using, say, a standard SQL technique), the latest stored data from a database, such as activation time, differential pressure, manifold pressure. In this example, the approach also retrieves data on the filter performance as a function of time from the latest obtained parameter sets. In some examples, the database may include other information / data, such as pollutant type, filter design, over the course of time analytics algorithms would define certain filter performance parameters, e.g., the rate and profile by which each filter becomes clogged over time, etc. The impact of this on overall cleaning efficiency (e.g., how well does the air filtration system return to the clean state), how many times can the air filtration system be cleaned before it degrades, etc. These (and other standard) parameters are used to provide benchmarks, typically set by the OEM, and based on historical data of what has always ‘worked’ previously. Examples described herein challenge this sole use of standard historical data approach.

[0123] At 710, a performance of a set of filters is compared based on a pollutant type and a filter design against a ‘best-in-class’ benchmark. At 715, a determination is made as to the whether (or not) recent parameter changes improve the performance of the filter system (with one or more filters). As parameters are adjusted, examples described herein monitor and analyse whether (or not) the existing benchmark is improved (or degraded) by the new (adjusted) parameters. Through this mechanism it is possible to determine whether (or not) the adjustment contributes to an improved filter performance. If an improvement is determined, then the adjusted parameters are set, which may be used to create a new benchmark for the type of air filtration system. It is also envisaged that these adjusted parameters may then also be used when designing future new installations.

[0124] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0125] - 19 - If the recent parameters do not improve the performance of the filter system at 715, the flowchart moves to 725. If the recent parameters do improve the performance of the filter system at 715, the flowchart moves to 720, where the benchmark parameters are updated with the current measured / estimated / determined parameters for a given pollutant type and filter design. Thereafter, at 725, a suggestion is provided for the latest benchmark parameters.

[0126] Referring now to FIG. 8, a yet further example of a stand-alone air filtration system 800 that captures manifold pressure to determine a pressure profile across each filter before, during and after a cleaning event is illustrated, in accordance with some examples. The yet further example of a stand-alone air filtration system 800 is similar in many respects to the abovedescribed air filtration system 200 of FIG. 2 and air filtration system 300 of FIG. 3, and therefore like functions and functionality will not be repeated here for the sake of clarity and to not obfuscate the explanations to a reader. The air filtration system 800 includes a compressed air manifold 810 that provides a manifold pressure sense measurement 820 to the analytical software processor 270. In some examples, the manifold pressure sense measurement 820 (and / or additional metrics) may be used to support improvements to the utilisation of compressed air as a mechanism for cleaning the filters, including at least one of: recommending an inter-clean cycle time, a recommended pneumatic manifold pressure recovery time, etc.

[0127] In examples described herein, the manifold pressure sense measurement 820 may perform the following tasks:

[0128] (i) It may provide a manifold pressure measurement at the time a valve is activated. This may be used to determine if the pressure is indeed sufficient to produce an effective clean of the filter. If the pressure is not high enough, then the cleaning cycle may be determined as not being the most efficient.

[0129] (ii) It may record a profile of the manifold pressure as the valve is open. This profile may be used to determine the impact that the blast of air is having on the filter. The impact has the potential to clean the filter but also damage the filter. Therefore, in accordance with some examples, using ML and Al it is possible to start to determine what would be the best profile for any given filter type that maximises cleaning performance, whilst reducing wear and tear, e.g., understand the trade-offs.

[0130] (iii) There is an expectation that there will be a certain level of manifold pressure drop whilst the valve is active. If the pressure drop does not follow a projected profile then

[0131] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0132] - 20 - manifold pressure sense measurement 820 may start to indicate a problem with the valve system, per se e.g. wear and tear on the valve, valve sealing issues or potential solenoid issues that required attention. This may help the supplier to accurately identify the exact location of a faulty system before attending to the issue on-site. This will save time, unnecessary call-outs, etc .

[0133] (iv) The manifold pressure sense measurement 820 may also be monitored, once the valve has closed and finished cleaning, in order to determine the rate at which the manifold pressure recovers. This pressure may be used to determine when the system is ready for a follow-on cleaning event, thereby enabling improved parameters to be set that will reduce compressor energy. This information may also be analysed to determine whether there are leaks in the system, since it is expected that the manifold would charge at a given rate, and remain at it’s peak level between cleaning events. If there are variances in these metrics, then it may indicate an issue with the air filtration system. In addition to using manifold pressure information to modify parameters for best efficiency of the air filtration system, it is envisaged that the manifold pressure information may also be used to diagnose potential faults with the system in the event a filter has not been cleaned properly, e.g., if there is solenoid current with no loss of manifold pressure then there is likely a fault with the valve.

[0134] The setting of the air pressure on the compressed air manifold is typically determined by one or both of the following criteria: the manufacturer will specify the maximum allowable pressure into the solenoid valve from the manifold, OEM / reseller may advise the min / max air pressure to the solenoid valve from the manifold.

[0135] Referring now to FIG. 9, a yet further alternative example of an integrated air filtration system 900 that captures manifold pressure to determine a pressure profile across each filter before, during and after a cleaning event is illustrated, in accordance with some examples. The yet further alternative example of an integrated air filtration system 900 is similar in many respects to the above-described stand-alone air filtration system 800 of FIG.

[0136] 8, and therefore like functions and functionality will not be repeated here for the sake of clarity and to not obfuscate the explanations to a reader. The integrated air filtration system 900 includes a compressed air manifold 910 that provides a manifold pressure sense measurement 915 to the communication controller 920. The communication controller 920 is modified to communicate the various parameters of the air filtration system 900 including the manifold pressure sense measurement 915 to an analytical software processor 270. In

[0137] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0138] - 21 -some examples, it is envisaged that combinations of integrated and stand-alone air filtration systems are supported, for example it is envisaged that an air filtration system may be integrated but have the manifold pressure feed directly into the intelligent cleaning controller or processor instead, for example using a wireless technology such as BLE™. In some examples, the manifold pressure sense measurement 920 (and / or additional metrics) may be used to support improvements to the utilisation of compressed air as a mechanism for cleaning the filters, including at least one of: recommending an inter-clean cycle time, a recommended pneumatic manifold pressure recovery time, etc.

[0139] Referring now to FIG. 10, a timing example of how manifold pressure is collated to determine a pressure profile of FIG. 8 or FIG. 9 in utilising compressed air for cleaning filters is illustrated, in accordance with some examples. The timing diagram 1000 includes a first waveform 1005 illustrating the operation of the first valve ‘AT with a first period of time S1 1010 that is ‘high’ when a short blast of compressed air is expelled into a first filter. The timing diagram 1000 includes a second waveform 1015 illustrating the operation of the second valve ‘A2’ with a second period of time S2 1020 that is ‘high’ when a short blast of compressed air is expelled into a second filter. The timing diagram 1000 includes a third waveform 1025 illustrating the operation of the third valve ‘A2’ with a third period of time Sn 1030 that is ‘high’ when a short blast of compressed air is expelled into the nth filter. A fourth waveform 1035 illustrates the consequent operation of the differential pressure from differential pressure sensor 268. As illustrated, the fourth waveform 1035 includes periods 1040, 1045, 1050 of reducing differential pressure that coincide with one of the solenoid valves being sequentially opened to short blasts of compressed air input to a respective filter. The timing diagram 1000 includes a fifth waveform 1055 of the manifold pneumatic pressure, as illustrated in FIG. 8 and FIG. 9. In alignment with the first valve ‘AT being activated at the first period of time S1 1010 that is ‘high’ a short blast of compressed air from the manifold is expelled into the first filter at 1060, with the short blast of compressed air stopping at 1065 at the end of the first period of time S1 1010. Similarly, when the second valve ‘A2’ is activated at the second period of time S2 1020 that is ‘high’ a short blast of compressed air from the manifold is expelled into the first filter at 1070, with the short blast of compressed air stopping at 1075 at the end of the second period of time S2 1020. In this manner, by monitoring the differential pressure and the valve opening periods, together with the sensed manifold pneumatic pressure described in FIG. 8 and FIG. 9, the controller or analytical software processor is able to determine and collate the filter differential pressure at the beginning of each valve’s cleaning event, the filter differential pressure at the end of

[0140] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0141] - 22 -each valve’s cleaning event, using the duration of the event based on, and obtained using, the valve activity signals. The addition of the manifold pressure measurements provides further insights to the impact of the cleaning of the filter. In some examples, these insights may be used to detect issues with a performance of the filter cleaning system, which may, in turn, be affecting the ability of the filter to clean the dirty environmental air.

[0142] Referring now to FIG. 11 , a yet further example of an air filtration system architecture 1100 that connects a cloud-based server that comprises a neural network and supports an automated feedback mechanism to rapidly adjust a cleaning operation of each filter is illustrated, in accordance with some examples. In some examples, it is envisaged that the automated feedback mechanism 1160 may use any suitable generic communication technology and provide real-time, automatic feedback of parametric suggestions for the analytical software processor 270 to apply as updates to the configuration of the air filtration system architecture 1100.

[0143] The air filtration system architecture 1100 illustrates the measurement and reporting 1115 of air filtration system parameters using any suitable communication technology / interface to a cloud-based server 530 that includes a cloud storage database 532. In this example, the cloud storage database 532 is configured to store one or more of the following: a filter pressure profile, a manifold pressure profile, valve activity, pollutant classification data, filter design parameters, etc.

[0144] In some examples, the cloud-based server 530 may be configured with neural network capability, as described with reference to FIG. 12, where data analytics 534 may be extracted from cloud storage database 532 and provided to a learning processor 538 that uses the neural network to provide air filtration recommendations to analytical software processor 270. In accordance with some examples, the cloud-based server 530 with the neural network is configured to learn from the data provided to it and continuously determine the most efficient system setup that maximises the filters' cleaning operations. In accordance with some examples, an automated feedback mechanism may be employed to rapidly adjust a cleaning operation of each filter, for example in a real-time manner. In some examples, the cloud-based server 530 may be made available to operators and management staff of the air filtration systems and configured to provide the latest suggested parameters to be setup for any filter system that includes analytical software processor 270, based on the pollutant types and filter design parameters. In some examples, the cloud-

[0145] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0146] - 23 -based server 530 may suggest parameters 1160 for each solenoid event including one or more of: valve duration, manifold recovery time, filter pressure thresholds, etc., as well as parameters for each solenoid event including one or more of: a sequence order of the solenoids, a repetition rate and sequence of each solenoid, a blast energy of each solenoid jet (pneumatic power x time), etc. In this example, the analytical software processor 270 also includes a filter / system configuration management 1120 function, which in some examples is implemented in a DSP, which may be a cloud-based processor.

[0147] Referring now to FIG. 12, one example of an operation of a neural network 1200 that may be employed as a learning processor, such as an artificial intelligence (Al)-based architecture configured to analyse air pressure in order to adjust a cleaning of at least one filter in an air filtration system, is illustrated according to some examples described herein. In some examples, the example neural network 1200 may be configured to suggest improvements that affect pneumatic blast energy as a function of valve activation time and / or manifold recovery times in order to improve compressor and filter energy utilisation. In some examples, the example neural network 1200 may be configured to perform failure prediction, for example based on reviewing patterns on how the differential pressure behaves during each cleaning event. For example, as filters wear out, the cleaning becomes less effective, and the pressure behaviour changes in a noticeable way. By comparing realtime data with previous good working examples, the air filtration system may be able to recognise when a filter may imminently fail and flag it for maintenance before it causes problems.

[0148] In some examples, the example neural network 1200 may comprise a convolutional neural network 1200, which applies a series of node mappings 1280 to an input 1210, which ultimately resolves into an output 1230 consisting of one or more values, from which at least one of the values is used by the a neural network 1200, for example the Al-based architecture of FIG. 5 or FIG. 11. The example convolutional neural network 1200 comprises a consecutive sequence of network layers (e.g. layers 1240), each of which consists of a series of channels 1250. The channels are further divided into input elements 1260. In this example, each input element 1260 may store a single value. Some (or all) input elements 1260 in an earlier layer are connected to the elements in a later layer by node mappings 1280, each with an associated weight. The collection of weights in the node mappings 1280, together, form the neural network model parameters 1247. For each node mapping 1280, the elements in the earlier layer are referred to as input elements 1260 and the elements in

[0149] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0150] - 24 -the output layer are referred to as the output elements 1270. An element may be an input element to more than one node mapping, but an element is only ever the output of one node mapping function 1220.

[0151] In order to calculate the output 1230 of the convolutional neural network 1200 the system first considers the input layer as the earlier layer. The layer(s) to which the earlier layer is connected by a node mapping function 1220 are considered in turn as the later layer. The value for each element in later layers is calculated using the node mapping function 1220 in equation [1], where the values in the input elements 1260 are multiplied by their associated weight in the node mapping function 1220 and summed together.

[0152] Node mapping function 1220: d = A(wadx a + wbdx b + wcdx c) [1]

[0153] The result of the summing operation is transformed by an activation function, ‘A’ and stored in the output element 1270. The convolutional neural network 1200 now treats the previously considered later layer(s) as the earlier layer, and the layers to which they are connected as the later layers. In this manner the convolutional neural network 1200 proceeds from the input layer 1240 until the value(s) in the output 1230 have been computed.

[0154] In examples herein-described, the convolutional neural network 1200 may be trained. In some examples of the invention, the training of the convolutional neural network 1200 may entail repeatedly presenting air filtration data as the input 1210 of the convolutional neural network 1200, in order to analyse the operation and / or efficiency of the individual filters in the reverse jet dust collector. In one envisaged example, initial training data may be the original default settings used by each supplier of the air filtration system when it is installed at particular types of industrial plants. Thereafter, when the air filtration system is installed, it is anticipated that the commissioning data may include information about the plant and filter design characteristics, for example. In some examples of the invention, an optimisation algorithm may be used to reduce a loss function, for example by measuring how much each node mapping 1280 weight contributed to the loss, and using this to modify the node mapping functions 1220 in such a way as to reduce the loss. Each such modification is referred to as an iteration. After a sufficient number of iterations the convolutional neural network 1200 can be used to analyse the operation and / or efficiency of the individual filters in the reverse jet dust collector from the input of air filtration data.

[0155] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0156] - 25 -ln some examples of the invention, the large number of model parameters 1247 used in the convolutional neural network may require the device to include a memory 1290. The memory 1290 may be used to store the training data 1215, the model parameters 1247, and any intermediate results 1293 of the node mappings.

[0157] Thus, in the air filtration system input data (a training dataset, clinical dataset, model parameters or intermediate results) is fed to the learning processor neuronal network in a format that fits the input matrix. Nodes are mapped in a specific way that is adapted to the purpose of the device (forming e.g. a convolutional neuronal network). The information is gradually reduced through a series of interconnected input I output elements to generate the final output. In some examples, it is envisaged that, depending on application needs, the learning processor / machine learning algorithms may be based on supervised, unsupervised, or reinforcement learning models.

[0158] Referring now to FIG. 13, a flowchart 1300 illustrates one example for adjusting filtration control parameters of an air filtration system and / or adjusting in a real-time manner at least one filter cleaning parameter. At 1305, the flowchart starts by receiving unfiltered air 262, by a reverse jet dust collector 250 that comprises an airflow inlet 260 and a plurality of filters 252, 254, 256. At 1310, the flowchart continues with applying pressurized air to clean individual filters 252, 254, 256 of the plurality of filters 252, 254, 256. At 1315, the flowchart continues with outputting filtered exhaust air 266 by an exhaust output 264 of the reverse jet dust collector 250. At 1320, the flowchart continues with monitoring a differential pressure between the airflow inlet 260 and the exhaust output 264 across the reverse jet dust collector 250 by a differential pressure sensor 268 whilst an individual filter 252, 254, 256 is being cleaned. At 1325, the flowchart continues with collating data that includes a differential pressure reading from the differential pressure sensor 268. At 1330, the flowchart continues with processing the collated data by a learning processor; and at 1335, the flowchart ends by adjusting at least one parameter of the pressurized air system in response to the processed collated data. Alternatively, at 1340, the flowchart ends by adjusting in a real-time manner at least one filter cleaning parameter in response to the processed data, wherein the cleaning parameter comprises at least one of: solenoid sequencing, a solenoid activation timing, energy of a pneumatic blast of filtered air.

[0159] In some optional examples, one of the analytical software processor (270) or learning processor may be configured to: track differential pressure as a function of solenoid

[0160] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0161] - 26 -activation time; and determine an effectiveness of filter cleaning on a per-solenoid, perfilter basis.

[0162] In some optional examples, the data provided to and collated by the analytical software processor (270) may further include an indication of an activation of individual solenoids and a timing of activation of individual solenoids. In some optional examples, the activation of individual solenoids may be used by the analytical software processor (270) as a timing basis for aligning filter activation in a time domain. In some optional examples, the activation of individual solenoids may be used by the analytical software processor (270) to track filter individual filter cleaning events in (substantially) real-time. A skilled artisan readily appreciates that reference to ‘real-time’ herein encompasses natural and unavoidable latency due to activation and processing of data, etc., and is intended to cover such real-life delays.

[0163] In some optional examples, the pressurised air system may be operably coupled to a compressed air manifold, where the processed collated data provided to and collated by the analytical software processor further comprises manifold pressure. Here, the analytical software processor is arranged, in response to the differential pressure reading and manifold pressure, to perform one or more of the following: provide an indication of an operational effectiveness of a valve; propose or execute a new parameter that adjusts a duration of time that the valve is active, propose or execute a new parameter that adjusts a duration of time needed to recharge the compressed air manifold; identify a cause of a valve problem, propose or execute a new parameter that adjusts valve sequencing. In this manner, in a design phase of a new filter system, a designer may be able to use the algorithms and results therefrom to help determine the optimum valve diameters for any given design. In some optional examples, one of the analytical software processor (270) or learning processor may be configured to: track differential pressure as a function of solenoid activation time and manifold pressure; and determine an effectiveness of filter cleaning on a per-solenoid, per-filter basis.

[0164] In an alternative aspect, an air filtration system is described that comprises: a reverse jet dust collector (250) that comprises an airflow inlet (260) to receive unfiltered air (262), a plurality of filters (252, 254, 256) arranged to filter air input into the airflow inlet (260) and an exhaust output (264) arranged to output filtered exhaust air (266); a pressurized air system configured to apply pressurized air to clean individual filters (252, 254, 256) of the plurality of filters (252, 254, 256); a differential pressure sensor (268) configured to monitor a differential pressure between the airflow inlet (260) and the exhaust output (264) across

[0165] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0166] - 27 -the reverse jet dust collector (250) whilst an individual filter (252, 254, 256) is being cleaned; an analytical software processor (270) operably coupled to the differential pressure sensor (268) and arranged to process received continuous air filtration system behaviour data that includes a differential pressure reading from the differential pressure sensor (268); and dynamically adjust in a real-time manner at least one filter cleaning parameter in response to the processed data, wherein the cleaning parameter comprises at least one of: solenoid sequencing, a solenoid activation timing, energy of a pneumatic blast of filtered air.

[0167] \ln an alternative aspect, an air filtration system is described that comprises: a reverse jet dust collector (250) that comprises an airflow inlet (260) to receive unfiltered air (262), a plurality of filters (252, 254, 256) arranged to filter air input into the airflow inlet (260) and an exhaust output (264) arranged to output filtered exhaust air (266); a pressurized air system configured to apply pressurized air to clean individual filters (252, 254, 256) of the plurality of filters (252, 254, 256); a differential pressure sensor (268) configured to monitor a differential pressure between the airflow inlet (260) and the exhaust output (264) across the reverse jet dust collector (250) whilst an individual filter (252, 254, 256) is being cleaned; an analytical software processor (270) is operably coupled to the differential pressure sensor (268) and arranged to collate data that includes a differential pressure reading from the differential pressure sensor (268); and a learning processor is operably coupled to the analytical software processor (270) and arranged to process the collated data and adjust at least one parameter of the pressurized air system in response to the processed collated data.

[0168] A skilled artisan will appreciate that the level of integration of circuits or components may be, in some instances, implementation-dependent. Clearly, the various circuits or components can be realized in discrete or integrated component form, with an ultimate structure therefore being an application-specific or design selection. It is envisaged that the analytical software processor concepts described herein may be applied to any system on chip (SoC) and any embedded processor. A skilled artisan will appreciate that the level of integration of components may be, in some instances, implementation-dependent. As the illustrated example embodiments may, for the most part, be implemented using electronic components and circuits known to those skilled in the art, details have not been explained in any greater extent than that considered necessary, as illustrated below, for the understanding and appreciation of the underlying concepts herein described and in order not to obfuscate or distract from the teachings presented.

[0169] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0170] - 28 -ln the foregoing specification, the examples have been described with reference to specific embodiments. It will, however, be evident that various modifications and changes may be made therein without departing from the scope as set forth in the appended claims and that the claims are not limited to the specific examples described above.

[0171] The connections as discussed herein may be any type of connection suitable to transfer signals from or to the respective nodes, units or devices, for example via intermediate devices. Accordingly, unless implied or stated otherwise, the connections may for example be direct connections or indirect connections. The connections may be illustrated or described in reference to being a single connection, a plurality of connections, unidirectional connections, or bidirectional connections. However, different embodiments may vary the implementation of the connections. For example, separate unidirectional connections may be used rather than bidirectional connections and vice versa. Also, plurality of connections may be replaced with a single connection that transfers multiple signals serially or in a time multiplexed manner. Likewise, single connections carrying multiple signals may be separated out into various different connections carrying subsets of these signals. Therefore, many options exist for transferring signals.

[0172] Those skilled in the art will recognize that the architectures depicted herein are merely exemplary, and that in fact many other architectures can be implemented which achieve the same functionality. Any arrangement of components to achieve the same functionality is effectively ‘associated’ such that the desired functionality is achieved. Hence, any two components herein combined to achieve a particular functionality can be seen as ‘associated with’ each other such that the desired functionality is achieved, irrespective of architectures or intermediary components. Likewise, any two components so associated can also be viewed as being ‘operably connected,’ or ‘operably coupled,’ to each other to achieve the desired functionality.

[0173] Furthermore, those skilled in the art will recognize that boundaries between the abovedescribed operations merely illustrative. The multiple operations may be combined into a single operation, where a single operation may be distributed in additional operations and operations may be executed at least partially overlapping in time. Moreover, alternative embodiments may include multiple instances of a particular operation, and the order of operations may be altered in various other embodiments. Also, for example, in one embodiment, the illustrated examples may be implemented as circuitry located on a single integrated circuit or within a same device. Alternatively, the circuit and / or component

[0174] Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final

[0175] - 29 -examples may be implemented as any number of separate integrated circuits or separate devices interconnected with each other in a suitable manner. Also, for example, the examples described herein, or portions thereof, may implemented as soft or code representations of physical circuitry or of logical representations convertible into physical circuitry, such as in a hardware description language of any appropriate type.

[0176] In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word ‘comprising’ does not exclude the presence of other elements or steps then those listed in a claim. Furthermore, the terms ‘a’ or ‘an,’ as used herein, are defined as one or more than one. Also, the use of introductory phrases such as ‘at least one’ and ‘one or more’ in the claims should not be construed to imply that the introduction of another claim element by the indefinite articles ‘a’ or ‘an’ limits any particular claim containing such introduced claim element to inventions containing only one such element, even when the same claim includes the introductory phrases ‘one or more’ or ‘at least one’ and indefinite articles such as ‘a’ or ‘an.’ The same holds true for the use of definite articles. Unless stated otherwise, terms such as ‘first’ and ‘second’ are used to arbitrarily distinguish between the elements such terms describe. Thus, these terms are not necessarily intended to indicate temporal or other prioritization of such elements. The mere fact that certain measures are recited in mutually different claims does not indicate that a combination of these measures cannot be used to advantage.

[0177] Optimus Confidential Proprietary

Claims

Attorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final- 30 - Claims1. An air filtration system comprises:a reverse jet dust collector (250) that comprises an airflow inlet (260) to receive unfiltered air (262), a plurality of filters (252, 254, 256) arranged to filter air input into the airflow inlet (260) and an exhaust output (264) arranged to output filtered exhaust air (266);a pressurized air system configured to apply pressurized air to clean individual filters (252, 254, 256) of the plurality of filters (252, 254, 256);a differential pressure sensor (268) configured to monitor a differential pressure between the airflow inlet (260) and the exhaust output (264) across the reverse jet dust collector (250) whilst an individual filter (252, 254, 256) is being cleaned;an analytical software processor (270) operably coupled to the differential pressure sensor (268) and arranged to collate data that includes a differential pressure reading from the differential pressure sensor (268); anda learning processor operably coupled to the analytical software processor (270) and arranged to process the collated data and adjust at least one parameter of the pressurized air system in response to the processed collated data.

2. The air filtration system of Claim 1 wherein the pressurized air system comprises a plurality of solenoids respectively operably coupled to the plurality of filters (252, 254, 256) and arranged to be individually activated to emit pressurized air into a respective filter (252, 254, 256), wherein the data provided to and collated by the analytical software processor (270) further comprises an indication of an activation of individual solenoids and a timing of activation of individual solenoids.

3. The air filtration system of Claim 1 or Claim 2 wherein one of the analytical software processor (270) or learning processor is configured to:track differential pressure as a function of solenoid activation time; and determine an effectiveness of filter cleaning on a per-solenoid, per-filter basis.

4. The air filtration system of Claim 2 wherein the activation of individual solenoids is used by the analytical software processor (270) as a timing basis for aligning filter activation in a time domain.

5. The air filtration system of Claim 2 wherein the activation of individual solenoids is used by the analytical software processor (270) to track individual filter cleaning events in real-time.Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final- 31 -6. The air filtration system of any preceding Claim wherein:the pressurized air system is operably coupled to a compressed air manifold, the processed collated data provided to and collated by the analytical software processor (270) further comprises manifold pressure; andthe analytical software processor (270) is arranged, in response to the differential pressure reading and manifold pressure, to perform one or more of the following:provide an indication of an operational effectiveness of a valve; propose or execute a new parameter that adjusts a duration of time that the valve is active,propose or execute a new parameter that adjusts a duration of time needed to recharge the compressed air manifold;identify a cause of a valve problem,propose or execute a new parameter that adjusts valve sequencing.

7. The air filtration system of Claim 6 wherein one of the analytical software processor (270) or learning processor is configured to:track differential pressure as a function of solenoid activation time and manifold pressure; anddetermine an effectiveness of filter cleaning on a per-solenoid, per-filter basis.

8. The air filtration system of any preceding Claim wherein the learning processor is further arranged to perform at least one of the following:predict at least one failure event of the pressurized air system;adjust a cleaning cycle of the plurality of filters (252, 254, 256) of the pressurized air system;adapt at least one component in the pressurized air system to affect an efficiency parameter of the air filtration system.

9. The air filtration system of any of preceding Claims 2 to 8 wherein the learning processor is further arranged to dynamically adjust at least one of the following: a solenoid activation time, a manifold recovery time, number of repeats on a valve, valve sequencing.

10. The air filtration system of any of preceding Claims 2 to 9 wherein the at least one parameter of the pressurized air system adjusted comprises at least one of: a cleaning parameter, solenoid sequencing, activation timing, pneumatic air blast energy.Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final- 32 - 11. The air filtration system of any preceding Claim wherein the learning processor is arranged to provide a real-time feedback to the pressurized air system based on continuous analysis of air filtration system behaviour to adjust, in response to the processed collated data, at least one of:a parameter of the pressurized air system.a cleaning parameter that comprises at least one of: solenoid sequencing, a solenoid activation timing, energy of a pneumatic blast of filtered air.

12. The air filtration system of any preceding Claim wherein the learning processor is arranged to capture and analyse differential pressure profiles associated with each cleaning event and assess individual filter performance.

13. The air filtration system of any preceding Claim wherein the learning processor is operably coupled to a cloud storage database (532) configured to store at least one of the following: a filter pressure profile, a manifold pressure profile, valve activity, pollutant classification data, a filter design parameters.

14. The air filtration system of any preceding Claim wherein the learning processor receives collated data from the analytical software processor (270) via at least one of: a wide area network communication system, a cellular communication system, an Internet of Things communication network, loT, a Bluetooth™ communication system, a Wi-Fi™ communication system.

15. The air filtration system of any preceding Claim wherein the learning processor, in response to the processed collated data, is configured to provide parameter recommendations when setting up the air filtration system based on data identifying at least: a pollutant type, a filter design, a filter size.

16. A method for adjusting filtration control parameters of an air filtration system, the method comprising:receiving unfiltered air (262), by a reverse jet dust collector (250) that comprises an airflow inlet (260) and a plurality of filters (252, 254, 256);applying pressurized air to clean individual filters (252, 254, 256) of the plurality of filters (252, 254, 256);outputting filtered exhaust air (266) by an exhaust output (264) of the reverse jet dust collector (250);Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final- 33 - monitoring a differential pressure between the airflow inlet (260) and the exhaust output (264) across the reverse jet dust collector (250) by a differential pressure sensor (268) whilst an individual filter (252, 254, 256) is being cleaned;collating data that includes a differential pressure reading from the differential pressure sensor (268);processing the collated data by a learning processor; andadjusting at least one parameter of the pressurized air system in response to the processed collated data.

17. An air filtration system comprises:a reverse jet dust collector (250) that comprises an airflow inlet (260) to receive unfiltered air (262), a plurality of filters (252, 254, 256) arranged to filter air input into the airflow inlet (260) and an exhaust output (264) arranged to output filtered exhaust air (266);a pressurized air system configured to apply pressurized air to clean individual filters (252, 254, 256) of the plurality of filters (252, 254, 256);a differential pressure sensor (268) configured to monitor a differential pressure between the airflow inlet (260) and the exhaust output (264) across the reverse jet dust collector (250) whilst an individual filter (252, 254, 256) is being cleaned;an analytical software processor (270) operably coupled to the differential pressure sensor (268) and arranged to:process received continuous air filtration system behaviour data that includes a differential pressure reading from the differential pressure sensor (268); and dynamically adjust in a real-time manner at least one filter cleaning parameter in response to the processed data.

18. The air filtration system of Claim 17, wherein the pressurized air system comprises a plurality of solenoids respectively operably coupled to the plurality of filters (252, 254, 256) and arranged to be individually activated to emit pressurized air into a respective filter (252, 254, 256), wherein the continuous air filtration system behaviour data further comprises an indication of an activation of individual solenoids and a timing of activation of individual solenoids.

19. The air filtration system of Claim 18, wherein the analytical software processor (270) is further arranged to use the timing of activation of individual solenoids in at least one of the following: as a timing basis for aligning filter activation in a time domain, to track individual filter cleaning events in real-time.Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final- 34 -20. The air filtration system of Claim 18 or Claim 19, wherein the at least one filter cleaning parameter comprises at least one of: solenoid sequencing, a solenoid activation timing, energy of a pneumatic blast of filtered air.

21. The air filtration system of any of preceding Claims 18 to 20 wherein the analytical software processor (270) is further arranged to:track differential pressure as a function of solenoid activation time; and determine an effectiveness of filter cleaning on a per-solenoid, per-filter basis.

22. The air filtration system of any of preceding Claims 17 to 21 wherein the pressurized air system is operably coupled to a compressed air manifold, and the continuous air filtration system behaviour data further comprises an air manifold pressure reading.

23. The air filtration system of Claim 22 when dependent upon Claim 18, wherein the analytical software processor (270) is arranged, in response to the differential pressure reading and air manifold pressure reading to perform one or more of the following:provide an indication of an operational effectiveness of a solenoid;propose or execute a new parameter that adjusts a duration of time that the solenoid is active;propose or execute a new parameter that adjusts a duration of time needed to recharge the compressed air manifold;identify a cause of a solenoid problem;propose or execute a new parameter that adjusts solenoid sequencing.

24. The air filtration system of Claim 22 or Claim 23, wherein the analytical software processor (270) is further arranged to dynamically adjust, in response to the continuous air filtration system behaviour data, at least one of the following: a solenoid activation time, a manifold recovery time, number of repeats on a valve, valve sequencing, a parameter of the pressurized air system.

25. The air filtration system of any of preceding Claims 22 to 24, wherein the analytical software processor (270) is configured to:track differential pressure as a function of solenoid activation time and manifold pressure; anddetermine an effectiveness of filter cleaning on a per-solenoid, per-filter basis.Optimus Confidential ProprietaryAttorney Docket No. CAM-2025-001 WO 21 -Jan-2026 Specification_Final- 35 - 26. The air filtration system of any of preceding Claims 17 to 25 wherein the analytical software processor (270) is further arranged to perform at least one of the following: predict at least one failure event of the pressurized air system;adjust a cleaning cycle of the plurality of filters (252, 254, 256) of the pressurized air system;adapt at least one component in the pressurized air system to affect an efficiency parameter of the air filtration system.

27. The air filtration system of any of preceding Claims 17 to 26 wherein the analytical software processor (270) is further arranged to process the received continuous air filtration system behaviour data to capture and analyse differential pressure profiles associated with each cleaning event and assess individual filter performance.

28. The air filtration system of any of preceding Claims 17 to 27 wherein the analytical software processor (270) is operably coupled to a cloud storage database (532) configured to store at least one of the following: a filter pressure profile, a manifold pressure profile, valve activity, pollutant classification data, a filter design parameters.

29. The air filtration system of any of preceding Claims 17 to 28 wherein the analytical software processor (270) is a learning processor arranged to employ machine learning algorithms.

30. A method for adjusting filtration control parameters of an air filtration system, the method comprising:receiving unfiltered air (262), by a reverse jet dust collector (250) that comprises an airflow inlet (260) and a plurality of filters (252, 254, 256);applying pressurized air to clean individual filters (252, 254, 256) of the plurality of filters (252, 254, 256);outputting filtered exhaust air (266) by an exhaust output (264) of the reverse jet dust collector (250);monitoring a differential pressure between the airflow inlet (260) and the exhaust output (264) across the reverse jet dust collector (250) by a differential pressure sensor (268) whilst an individual filter (252, 254, 256) is being cleaned;processing received continuous air filtration system behaviour data that includes a differential pressure reading from the differential pressure sensor (268); and dynamically adjusting in a real-time manner at least one filter cleaning parameter in response to the processed data.Optimus Confidential Proprietary

Images