Computer-implemented method and system

Abstract

A computer-implemented method is provided which enables a user to utilise historical data to optimise the configuration to a drying environment.

Description

FIELD

[0001] The invention relates to a method and system. Particularly, but not exclusively, the invention relates to a computer-implemented method and system. Particularly, but not exclusively, the invention relates to a computer-implemented method of configuring a drying environment for the drying of plasterboard.BACKGROUND

[0002] Plasterboard is a widely used building material. When used in construction, plasterboard assist builders and designers in the control of condensation, fire protection and insulation (both thermal and acoustic). Plasterboard is usually used to make internal walls and ceilings.

[0003] Plasterboard comprises a core between two facers. The facers are usually either heavy duty construction paper or glass fibre mats where the glass fibre mats may be fully or partially embedded in the gypsum. The core comprises mostly calcium sulfate dihydrate, also known as gypsum. The plasterboard is formed depositing a wet slurry onto a first facer, before adding the second facer on top. The slurry comprises water, calcium sulfate hemihydrate, also known as stucco, and other additives such as foam and setting accelerators. The calcium sulfate hemihydrate in the slurry hydrates to form calcium sulfate dihydrate, but there is more water in the slurry than is needed to fully hydrate all of the calcium sulfate hemihydrate, therefore this excess water must be removed from the board by drying.

[0004] A large amount of free water cannot be left in a board, since this will make it heavy and cause it to degrade with time. However, boards must be dried carefully, since exposing the gypsum to temperatures which are too high can dehydrate the gypsum, producing calcium sulfate anhydrite, which will result in a board with unacceptable mechanical, acoustic and thermal insulating properties. In addition, the energy required to dry plasterboards accounts for a large proportion of the total energy demands and CO2 footprint, so maximising efficiency is key. Therefore, the drying step is an important part of the manufacturing process and the control of the temperature profile through the dryers is a complex.

[0005] Aspects and embodiments were conceived with the foregoing in mind.SUMMARY

[0006] Aspects relate to the drying of plasterboard. However, aspects may also be applied to the drying of other material products.

[0007] Viewed from a first aspect, there is provided a computer-implemented method of generating configuration data which can be used to configure a drying environment for the drying of plasterboard. The configuration data comprises one or more parameters which can be used to control aspects of a drying environment. Example aspects may comprise settings for components of a drying environment. A drying environment may comprise one or more zones. Example settings may include one of more of temperature settings, fan settings, exhaust settings, air pressure settings, waste temperature settings, humidity settings or damper settings for one or more zones. Settings may also relate to line speeds and the weights or relative weights of components which are used in the plasterboard production process.

[0008] The method may be implemented by a processing resource. The processing resource may comprise one or more processors. The processing resource may be implemented using software or hardware or a mixture of both.

[0009] The method may comprise obtaining a first set of historical data associated with the drying of plasterboard, wherein the historical data comprises data associated with energy consumption measurements for the drying of plasterboard. The association between data and energy consumption may be determined based on an increase or decrease in energy consumption caused by a parameter which is contained within the data.

[0010] The data may be categorised into parameters related to the drying of plasterboard. Example parameters may include parameters associated with one or more of fan speeds, temperatures, air pressures, exhaust pressures and the number of plasterboard substrates which are dried at one time inside a drying environment. Other example parameters may include or be associated with one or more of air humidity, air temperatures inside the zones, fresh air temperature, heat exchanger pressure discharge, air pressures at zone inlet and outlet, upstream parameters, dampers position, gas ratio between zones, residual moisture sensor, dryer filling level, line speed, the type and amount of fuel used, usage of auxiliary energy sources and dryer load. The energy consumption measurements may comprise measurements of the usage of one or more of electricity, gas or water. The historical data may comprise one or more parameters which are classified as categorical data. The historical data may comprise measurements associated with one or more parameters of the drying environment used in the drying of plasterboard. The historical data may be filtered based on required production parameters such as required moisture levels, line speeds and other drying environment settings. The historical data may be obtained responsive to input via a user interface. The user interface may be provided using any suitable computing device.

[0011] The method may further comprise using the historical data to identify parameters which are responsible for increases or decreases in energy consumption in the process of drying of plasterboard. Alternatively or additionally, other target variables such as, for example, moisture level may be used as the target variable in the processing of the historical data. Using historical data to identify parameters which increase or decrease energy consumption may comprise applying a trained model to the historical data to identify the parameters which correspond with an increase or decrease in energy consumption. Identification of parameters responsible for increases or decreases in energy consumption may comprise calculation of a correlation between the respective quantities. The correlation between a parameter and energy consumption may be determined by calculating the correlation coefficient between the respective parameter and the energy consumption. This may be determined analytically or numerically. The identification of parameters which correlate with energy consumption may comprise identification of parameters and data values in the historical data which minimises energy consumption. The parameters may be identified by the training of a trained model or the training and testing of a trained model. The trained model may be a regression-based model which relates the energy consumption to one or more parameters in the historical data.

[0012] The method may further comprise providing an output data set using the trained model, the output data set identifying parameters which minimise energy consumption in the drying of plasterboard. The output data may specify values of the parameters which are associated with minimised energy consumption. The output data may specify values associated with maximising or minimising another target variable e.g. moisture level. The output data set may comprise a configuration file provided in any suitable format. The format may be selected based on the operating or control system being used in a drying environment.

[0013] The trained model may be used to predict the parameters which will result in a drying environment which uses a minimal amount of energy. This may be by using the trained model to test values found in the historical data to determine the impact of those variables on energy consumption.

[0014] A method in accordance with the first aspect enables a drying environment to be configured in a way which minimises energy consumption. This is based on predictions obtained using historical data associated with the plasterboard drying process.

[0015] Any one of the steps of the method in accordance with the first aspect may be executed as a background process in that it runs without user intervention. This means that a drying environment may be configured in such a way that it minimises energy consumption but without the user to onerously need to select specific combinations of parameters which enable the optimal configuration to be generated.

[0016] In other words, a method in accordance with the first aspect enables a drying environment to be configured with minimal user interaction. The configuration is optimised to enable a drying environment to maintain a required level of quality in the plasterboard (e.g. a required moisture level) whilst minimising the energy consumption.

[0017] The method may further comprise providing a user interface displaying the parameters which minimise energy consumption in the drying of plasterboard. The user interface may be provided using any suitable computing device.

[0018] The method may comprise configuring a drying environment using the configuration set by the output data set. That is to say, the configuration of the drying environment is set by the output data set.

[0019] Optionally, the method may further comprise generating a configuration file for the drying environment based on the parameters which minimise energy consumption. The configuration file may comprise settings for one or more of the components of the drying environment being configured. Examples include fan speeds, air pressures, heating settings and the water pressure used in any water-based aspects of the drying environment.

[0020] The configuration file may comprise parameters identified by the trained model as being associated with minimisation of energy consumption during the drying of plasterboard.

[0021] The configuration file may be transmitted to the drying environment. The configuration file may then be used to configure the drying environment in that the settings characterised by the parameters are used to implement the drying environment.

[0022] Optionally, the configuration file may be generated responsive to user input requesting the configuration file. The user input may be provided via a user interface.

[0023] The configuration file may be generated responsive to user input selecting parameters for the drying environment using the user interface. The parameters may comprise at least one of average board moisture, line speed of the drying environment and the dryer filler. This enables the configuration file to be generated with the inclusion of the required production parameters. Other parameters may include data relating to specific time periods within the historical data.

[0024] Dryer filler can be defined as the proportion of the dryer section which is occupied. In other words, it defines the percentage of the maximum capacity of the dryer section which is occupied at a given time.

[0025] The user interface may display the parameters predicted by the optimisation as a range between minimum values and maximum values for respective parameters. For example, the minimum values and maximum values may define ranges for settings of the drying environment. The user interface may provide a selectable link which enables the configuration comprising the predicted parameters to be provided to the drying environment.

[0026] The predicted parameters are identified by the trained model and may be identified via the user interface using indicia to identify parameters associated with an optimised configuration.

[0027] The indicia may comprise lines, colours, words or other alphanumeric characters. The gain provided by selecting one parameter relative to another may also be displayed using such indicia.

[0028] The user interface may display positive and negative correlations associated with parameters. The user interface may display the positive and negative correlations as Shapley Additive Explanations (SHAP values) which indicate whether a parameter of the drying environment has a positive or negative correlation with the energy consumption of a drying environment. The positive and negative correlations may be determined during the identification of the parameters to identify associations between parameters of the historical data and the consumption of energy. A positive correlation may indicate that a parameter of the historical data is responsible for an increase in energy consumption. A negative correlation may indicate that a parameter of the historical data is responsible for a decrease in energy consumption.

[0029] Optionally, the user interface may provide selectable links to enable the user to select the configuration which minimises energy consumption.

[0030] Using the historical data to identify parameters which correlate with energy consumption in the process of drying of plasterboard may comprise training a regression model using the historical data to identify optimised hyperparameters. This may comprise utilising a search over the set of hyperparameters to identify a hyperparameter. The search may comprise using a grid search to search for an optimised hyperparameter. The grid search may utilise an R2 metric. Using the historical data to identify parameters which correlate with energy consumption may comprise testing the model using the historical data to produce a trained model. This may comprise evaluating the model using the historical data and optimised hyperparameters. The testing may comprise modifying the hyperparameters depending on the evaluation of the model.

[0031] The regression model may be applied as part of an artificial neural network (ANN). ANNs are otherwise known as connectionist systems which are computing systems which are vaguely inspired by biological neural networks. Such systems "learn" tasks by considering examples, generally without task-specific programming. They do this without any prior knowledge about the task or tasks, and instead, they evolve their own set of relevant characteristics from the learning / training material that they process. ANNs are considered nonlinear statistical data modelling tools where the complex relationship between inputs and outputs are modelled or patterns are found. An ANN may be hardware (where neurons are represented by physical components) or software-based (computer models) and can use a variety of topologies and learning algorithms. It will be appreciated that other feedforward models may also be used.

[0032] In an embodiment, the method may further comprise training a neural network model on training data to obtain the trained neural network model. Training data may comprise historical data associated with the plasterboard drying process. In an embodiment, said training may comprise applying a stochastic gradient descent algorithm on said training data. In an embodiment, the neural network may be trained using reinforcement learning.

[0033] The training of the regression model may comprise a first step wherein a first subset of the historical data is used to optimise the hyperparameters of the model and a second step wherein a second subset of the historical data is used to test the model. The first subset and the second subset may be identified based on time periods which define when the historical data was captured. Prior to the training of the regression model, the historical data may be pre-processed to filter out data which is not consistent with the production parameters identified by the user. The second subset may comprise historical data taken during a specified time period.

[0034] The trained model may be a categorical boosting (CatBoost) regression based model. This enables the model to handle categorical data such as, for example, set quality levels or identifications of drying environment.

[0035] The energy consumption may be associated with at least one of electricity, water or gas consumption.

[0036] In a further aspect, there may be provided a dryer for the drying of plasterboard. The drying environment may comprise a plasterboard feed component and a plurality of zones each configured to receive a quantity of plasterboard. Each of the plasterboard feed component and the plurality of zones may be configured using a method in accordance with the first aspect.

[0037] Further aspects may provide a system configured to implement the method of the first aspect and computer-program product which configures a processing medium to implement the steps of the first aspect.

[0038] Further aspects may provide a non-transitory storage medium configured to store instructions which, when executed by suitably configured hardware, provides instructions to a processing medium to implement the steps of the first aspect.DESCRIPTION

[0039] An embodiment is now described by way of example only and with reference to the following drawings in which: Figure 1 illustrates a processing resource configured to implement a method in accordance with the embodiment; Figure 2 illustrates the steps of a method in accordance with the embodiment; Figure 3 illustrates the steps of a method in accordance with the embodiment; Figures 4, 5a, 5b and 6 illustrate examples of user interfaces which could be provided to a user; Figure 7 illustrates an example of a zone of a longitudinal drying environment which may be configured in accordance with the embodiment; Figure 7a illustrates a longitudinal drying environment; Figure 8 illustrates a transversal drying environment which may be configured in accordance with the embodiment; and Figure 9 schematically illustrates the interaction between a drying environment interface and the control system of a drying environment.

[0040] We will now illustrate, with reference to Figure 1, a processing resource 100 configured to implement a method in accordance with the embodiment.

[0041] The processing resource 100 comprises a regression training module 102, a regression testing module 104 and a configuration generation module 106.

[0042] The regression training module 102 is configured to retrieve data from a datastore 112. The regression training module 102 is also configured to receive input via a user interface which will be described later on. The datastore 112 may be implemented using any suitable data storage. The data storage may be implemented using suitable hardware or software. The retrieval of data from the datastore may be implemented using any suitable technique. The communication of the data may be implemented using any telecommunications media.

[0043] The datastore 112 stores historical data which is collected from the plasterboard drying process. The historical data may be stored as a database which may be managed by a database management system. The historical data comprises fields which are associated with the plasterboard drying environment. The historical data further comprises fields which are associated with energy consumption of the drying process. The data associated with energy consumption (or usage) may comprise data which contains measurements of at least one of electricity, gas or water usage during the drying process. The historical data may be categorised into parameters related to the drying of plasterboard. Example parameters may include parameters associated with one or more of fan speeds, temperatures, air pressures, exhaust pressures and the number of plasterboard substrates which are dried at one time inside a drying environment. Other example parameters may include or be associated with one or more of air humidity, air temperatures inside the zones, fresh air temperature, heat exchanger pressure discharge, air pressures at zone inlet and outlet, upstream parameters, dampers position, gas ratio between zones, residual moisture sensor, dryer filling level, line speed, the type and amount of fuel used, usage of auxiliary energy sources and dryer load

[0044] The regression testing module 104 is configured to receive data from regression training module 102 and data from the datastore 112.

[0045] The configuration generation module 106 is configured to receive data from regression testing module 104 and is further configured to output data to a user interface module 108.

[0046] The user interface module 108 is configured to generate configuration files and to send those configuration files to a drying environment interface 110 which sets parameters for a plasterboard drying environment. The user interface module 108 is configured to receive user input via a user interface provided on a display of a computer system. The computer system may comprise any suitable computer such as, for example, a laptop computer, a desktop computer or a mobile telephone.

[0047] The data communications between the modules of the processing resource 100 may be enabled using any suitable data communications technique.

[0048] The processing resource 100 may be implemented using any suitable hardware or software. The processing resource 100 may be cloud implemented. The modules may be located remotely relative to each other.

[0049] We will now describe, with reference to Figure 2, how a drying environment can be configured to minimise energy consumption using the processing resource 100.

[0050] In a step 200, a user interface is initialised using the user interface module 108 and displayed on a computer operated by a user. The user selects the features which are to be used as the basis for the training and testing process. The features selected are used as the basis of modelling the energy consumption of the drying process, whether that be electricity consumption, energy consumption or water consumption. The user interface will only accept inputs related to the parameters which can be used to configure a drying environment. Some parameters will be restricted from user input, i.e. they are restricted features of the drying environment. The restricted features may be set by the physical configuration of the hardware used in the drying process e.g. the number of zones. The restricted features may also be set by safety concerns e.g. temperature settings or line speed settings.

[0051] More specifically, the drying sections used in the plasterboard manufacture process are made up of multiple different zones which all need to be configured in order to enable the dryer section to dry the plasterboard to a required level of quality (i.e. moisture content). Plasterboard is sensitive to the temperature and humidity profile of each zone and multiple dryer zones allow more precise control of the drying process as each zone can be configured separately. A "prezone" may also be included within a dryer section where warm air may be passed over the plasterboard at a lower temperature than is used in the proceeding zones. It may be that the prezone is heated by warm exhausted air from other zones.

[0052] Air is exhausted from each zone by an exhaust which typically comprises a fan moving at a speed which is configured by the operator.

[0053] The parameters which can be used to configure a drying environment may include parameters of the filters of the dryer sections (e.g. the type, the thickness and the width). The parameters may include settings on minimum and maximum dryer filler for each of the zones. The parameters may include settings on minimum and maximum dryer load. The parameters may include settings on minimum and maximum line speed for each zone. The parameters may include settings on minimum and maximum average humidity for each zone. The present price of electricity, gas and water may also be entered. Other parameters may include one or more of the following (for an example of a three zone dryer section): Table 1: Drying Environment ParametersParameterPhysical MeaningDryer_deck_speed_meanThis parameter (in metres / second) describes the mean average speed of movement the decks on which the plasterboard is laid during the drying of the plasterboard. Moving the plasterboard too quickly through each zone will mean that the plasterboard will not dry properly. Moving the plasterboard too slowly through each zone will dry too quickly and the moisture level will be too level.Dryer_speed_meanThis parameter (in metres / second) describes the speed at which the plasterboard progresses through each zone.Dryer_wet_deck_speed_meanThis parameter (in metres / second) describes the speed at which stucco ismoved into the pre-zone of the dryer environment, i.e. the warm zone where the boards are initially placed in the dryer environment.Dryer_Zone1_air_temperature_outlet_meanThe mean temperature of the air (in Celsius) which is being exhausted from zone 1 during the drying stage.Dryer_Zone2_air_temperature_inlet_meanThe mean temperature of the air (in Celsius) which is being introduced into zone 2 during the drying stage.Dryer_Zone2_delta_t_meanThe average temperature gradient across Zone 2.Dryer_Zone3_air_temperature_inlet_meanThe mean temperature of the air (in Celsius) which is being introduced into zone 3 during the drying stage.Dryer_Zone3_delta_t_meanThe average temperature gradient across Zone 3.Dryer_Zone3_setting_mirror_SP_meanSP is the set point of the temperature in Zone 3 and this corresponds to the mean temperature in zone 3. The setting mirror provides a factor by which the temperature in zone 3 can be shifted to accommodate for error in the calculation.Mixer_water_gaugeThis parameter describes the ratio of water to plaster (in litres / kilogram) in the slurry used to make the plasterboard.Dryer_load_ton_h_meanThe mean average amount of plasterboard in the dryer section (in kilograms)Dryer_waste_pressure_pv_meanThe pressure of the waste outlet which directs the waste material out of the drying section. It is expressed as a change in volume (g / m 3< ) and pressure (in Pascals)Dryer_Zone1_air_temperature_inlet_meanThe mean temperature of the air (in Celsius) which is being introduced into zone 1 during the drying stage.Dryer_Zone1_delta_t_meanThe average temperature gradient across Zone 1.Dryer_Zone2_air_temperature_middle_meanThe mean average temperature of the air at the middle of zone 2 of the dryer section.Dryer_Zone2_filling_level_meanThe amount of plasterboard (expressed as a number of decks) which is in zone 2 of the dryer section.Dryer_Zone3_air_temperature_middle_meanThe mean average temperature of the air at the middle of zone 3 of the dryer section.Dryer_Zone3_filling_level_meanThe mean average amount of plasterboard (expressed as a number of decks) which is in zone 3 of the dryer section. The average is determined over time (i.e. per second).Dryer_Zone3_settings_Mirror_OA_meanCompensation factor for error in temperature setting (determined by temperature gradient).Mixer_wet_weightThe weight (in grams) of the stucco introduced into the drying section.Dryer_prezone_pressure_meanThe mean air pressure (in Pascals) of the air in the prezone of the drying section.Dryer_waste_temperature_meanThe mean temperature (in Celsius) of the waste material being taken out of the drying section.Dryer_Zone1_air_temperature_middle_meanThe mean average temperature (in Celsius) of the air at the middle of zone 1 of the dryer section.Dryer_Zone1_filling_level_meanThe mean average amount of plasterboard (expressed as a number of decks) which is in zone 3 of the dryer section. The average is determined over time (i.e. per second).Dryer_Zone2_air_temperature_outlet_meanThe mean temperature of the air (in Celsius) which is being exhausted from zone 2 during the drying stage.Dryer_Zone2_outlet_pressure_meanThe mean air pressure (in Pascals) for the air exhausted from Zone 2.Dryer_Zone3_air_temperature_outlet_meanThe mean temperature of the air (in Celsius) which is being exhausted from zone 3 during the drying stage.Dryer_Zone3_outlet_pressure_meanThe mean air pressure (in Pascals) for the air exhausted from Zone 3.Line_speed_meanThe mean speed (in metres / second) at which the plasterboard is moved through the entire dryer section by the conveyor. A higher line speed means increased numbers of boards can enter the dryer per minute.Dryer_Zone3_exhaust_air_fan_meanThe mean average speed of the fan of the exhaust of Zone 3.Dryer_Zone3_fresh_air_damper_meanThe mean average position of the blades of the fresh air damper in Zone 3, i.e. the position of the blades which are used to regulate the flow of fresh air into Zone 3.Dryer_Zone3_air_humidity_meanThe mean average humidity in Zone 3.Dryer_prezone_exhaust_fan_speed_meanThe mean average speed of the fan of the exhaust in the prezone of the dryer section.Dryer_Zone1_air_humidity_meanThe mean average humidity in Zone 1.Dryer_Zone2_air_humidity_meanThe mean average humidity in Zone 2.Dryer_Zone2_fresh_air_damper_meanThe mean average position of the blades of the fresh air damper in Zone 2, i.e. the position of the blades which are used to regulate the flow of fresh air into Zone 2.

[0054] Even if the drying environment parameters cannot be input by the user at step S200 because they are restricted, the methods described herein may provide output settings for those parameters which can then be used to configure a drying environment.

[0055] Alternatively or additionally, even drying environment parameters which are not input at step S200 may be part of the configuration and may be output by the methods described herein.

[0056] These parameters are described in the context of a 3 zone dryer section but it should be understood that the parameters can be extended over any number of zones and that these parameters can be set using the user interface in step S200.

[0057] Generally speaking, the user interface may well be used to set parameters related to at least one or more of the exhaust of each zone, the air pressure in each zone, the air temperature in each zone, the waste temperature in the dryer section, the humidity in each zone and the position of the blades of dampers used in each zone. Some of the parameters may be restricted completely or restricted to specific ranges. The restrictions may be operator defined or manufacturer defined. Parameters may also be set which compensate for temperature difference in each zone.

[0058] Alternatively or additionally, any one of these parameters may be restricted in that they cannot be set by a user. This may be because the particular dryer environment does not admit adjustment of that parameter.

[0059] The mixer water gauge corresponds, as set out above, to the ratio of water to plaster. Ideally, only the water needed to hydrate stucco into calcium sulfate dihydrate. However, we need to have a slurry which can be poured in order to form a plasterboard precursor and therefore extra water is often added (it could be 0.6-1.2 times the weight of the stucco). The excess water, i.e. the water which does not become chemically bound within the dihydrate that the drying of the plasterboard needs to remove.

[0060] The mixer wet weight may also be a parameter which is input using the user interface. The mixer wet weight corresponds to the amount of water (by % compared to the plaster weight) which will be used to hydrate the calcium sulfate hemihydrate. In contrast, the dryer load is a parameter which can also be input which represents the amount of excess water that needs to be removed during the drying process.

[0061] In step S202, the user interface receives a selection of the energy consumption parameter which is to be minimised in the process of training and testing the model as described below. The energy consumption parameter may be one of electrical energy, gas or water. That is to say, the user selects a target parameter for the model training and testing process. Additional information may also be sourced at this stage, such as, for example, the calorific value of the gas or even unit costs of the electrical energy, gas or water. The target parameter may be defined in terms of watts (i.e. to minimise electrical energy usage), cubic feet or cubic metres (i.e. to minimise gas usage) or litres (i.e. to minimise water usage). The target parameter may also be line-speed. The target parameter may also define acceptable quality levels for the produced plasterboard (expressed in terms of moisture level) such as, for example, dryer load and the dryer zone filling for each of these products. Other target parameters may include one or more of a burning indicator, hemihydrate levels, end burnt or paper adhesion. The board temperature may also be used as a target parameter if an infrared sensor is used at the end of the drying stage. For the sake of this illustrative example, the selected energy consumption parameter is electrical energy and this is the target parameter for the example model training and testing process. Overall energy consumption may also be selected and this will focus on the consumption of both electricity and gas during the drying process. The selection of electrical energy should be taken to be limiting.

[0062] The user interface may receive selection of the specification of the drying environment and parameters which define boundaries on the average board moisture, line speed, dryer filling etc.

[0063] In step S204, the user interface receives a selection of parameters which define boundaries on the production of plasterboard using the drying environment which is being configured. These parameters may be described as stable production parameters. This allows the user to specify the quality requirements of the output of the drying environment, i.e. the quality of the plasterboard which has come through the configured drying environment. The boundary parameters selected in this step may be one or more of average board moisture, line speed or dryer filling. Step S204 may also enable the user to omit some parameter ranges from consideration, e.g. temperatures which are going to be too high for a specific drying environment or moisture levels which are going to be too low for a specific application of the plasterboard. The user may then initialise via the user interface the optimisation through the user interface.

[0064] In step S206, the regression training module 102 automatically (i.e. without user input) retrieves historical data from the datastore 112. The historical data contains values which are taken during the plasterboard drying process for previously produced plasterboard.

[0065] The historical data may be arranged as, for example, a Microsoft Excel file with 4 spreadsheets which are: Data Information, Linkage, Linkage Example and Product list table.

[0066] The Data Information may comprise fields relating to the plasterboard type, plasterboard thickness, plasterboard width, the production line speed, the external temperature, the wet weight of the plasterboard, the dry weight of the plasterboard, the water to evaporate, dryer load, water gauge, Gas / Oil flow measurements, the Gas PCS, the electricity measurements, the electricity measurements for the heating of the boiler, the dryer speed, the zone filling level and the air temperature.

[0067] The Linkage spreadsheet contains two sections. The first section relates to the dryer specification and the second section relates to the details regarding the belts, the wet end transfer, the speeds of infeed drives, the lengths and overspeed of the zones of the specified dryer. The overspeed of the zones of the specified dryer can be defined as the line speed of a zone by reference to the line speed of another zone. For example, a first zone may have a speed A and a second zone may have an overspeed defined as A + / - 5% to indicate it is 5% higher or lower than the first zone.

[0068] The Product list spreadsheet provides the information on the different products which have been produced in the respective plant. This spreadsheet contains standard value ranges of line speeds and acceptable quality levels for the produced plasterboard (e.g. expressed in terms of moisture level), dryer load and the dryer zone filling for each of these products. This spreadsheet may also specify product code, thickness, width and the number of rafts used side by side inside the dryer.

[0069] The historical data may be pre-processed to determine that all necessary information has been provided. This may be implemented using standard Python routines in the PreProcessing class such as, for example, read_raw_tables, pre-processing and run_preprocessing routines. The Linkage spreadsheet can then be used to generate a Linkage table. Information such as, reference step, line speed, distance between each step of the drying environment, time delay across the drying environment. For example if the reference step is the transfer to takeoff time of the drying environment, we enter the time when the board is at the beginning of each zone in the drying environment for specific drying configurations.

[0070] The historical data may be separated using an identifier such as, for example, a fabrication number. The regression training module 102 may be configured to automatically apply various data processing techniques to the historical data retrieved from the datastore 112. These techniques may include one or more of filtering, normalisation or removal of at least part of the retrieved data. For example, the historical data may be filtered to eliminate any data which would not satisfy the stable production conditions specified in step S204. For example, data related to plasterboard which was of an average board moisture which was higher or lower than the average board moisture specified in the stable production parameters provided in step S204 may be omitted from the historical data used by the regression training module 102. This is step S208. Alternatively or additionally, tolerances may be specified which provide boundaries on the stable production parameters and only data related to plasterboard which does not fall within those stable production parameters may be removed from the historical data used by the regression training module 102. If the regression training module cannot identify data which satisfies the stable production parameters, then it may be configured to search for data which can represent a good starting point such as data relating to produced plasterboard which has an average board moisture within a specified range of the value specified by the user in step S204.

[0071] This means that the trained model is trained on historical data which corresponds with plasterboard which has satisfied the stable production parameters which are specified by the user in step S204. This means that the configuration of the drying environment which is generated will be for the drying of plasterboard which matches the quality requirements required by the user. It may also mean that historical data for drying environments where the line speed, for instance, cannot be provided by the instant drying environment is also removed and will not feed into the predictions generated by the trained model.

[0072] The historical data may also specify that the plasterboard was dried using a specific drying configuration which is distinct from the one being configured by the user. For example, the instant drying environment may utilise a transversal dryer. The step in S208 may then remove all data associated with plasterboard which was dried using a longitudinal dryer, for example.

[0073] The step in S208 may utilise the feature_extraction module in Python to ensure the features are pulled from the data in a form suitable for further processing in other Python routines. The feature_extraction module utilises the FeatExtractor class which utilises some general routines such as, for example, common_features and get_features. This enables the data corresponding to the features of interest to be extracted. It can also be used to import plant dependent features which need to be taken into consideration in the modelling when configuring a drying environment in a specific manufacturing plant, e.g. number of zones, decks and rafts.

[0074] The processing of the data may be repeated over several iterations if, for example, more than one stable production parameter is specified so as to remove any data which will unnecessarily bias the training by inclusion of data which is not going to contribute to the optimisation of the desired drying environment. The iterations may be hierarchical in that, in the first iteration, a higher priority stable production parameter is used (e.g. a specific moisture level) and then successive lower priority variables (e.g. number of zones) may also be used as the basis for the removal of data.

[0075] On retrieval of the historical data, and any subsequent processing (including filtering), the regression training module 102 automatically initialises the training of a regression model which will predict the parameters of the historical data which will have a positive or negative correlation with the selected target parameter, i.e. which of electricity, gas or water is being minimised. This is step S210. An example of such a regression model is a categorical boosting (CatBoost) regression model as will be described below.

[0076] In a step S212, the regression training module 102 separates the processed historical data into a training set and a test set. The separation may be implemented using any suitable criteria. For example, the training set may be selected as 80% of the processed historical data and the test set may be selected as 20% of the processed historical data. In another example, the training set may correspond to data which was generated before a specified date and the test set may correspond to data which was generated on or after the specified date. The test set, for example, may correspond only to the data corresponding to the previous month's plasterboard production. In another example, the fabrication number of the historical data may be used as a grouping characteristic to group the historical data into a set of training data and a set of testing data. In the example where the fabrication number is used as the grouping characteristic, the test data set may comprise the data corresponding to the fabrication numbers for the previous month.

[0077] The regression training module 102 may, alternatively or additionally, be configured to train the regression model as part of the training of an artificial neural network (ANN). The artificial neural network may be trained using the historical data where the input comprises the stable production parameters and the output is the parameters of the dryer configuration. The hidden layers apply weights and biases which are optimised during the training process by the application of feedforward propagation where the ANN is trained to predict a configuration based on stable production parameters. The training may incorporate data associated with a specific drying environment e.g. longitudinal or transversal, number of dryer zones etc.

[0078] The test data set may, alternatively or additionally, correspond to the data corresponding to the final production of plasterboard for each month. This way of splitting data can ensure representation of data seasonality in the test set. The effect of this is that the seasonal changes in outside temperature and humidity can be incorporated into the data. Alternatively or additionally, the test data set may, alternatively or additionally, correspond to the data corresponding to a first production of plasterboard for each month. That is to say, the test data set may be specified by a time interval so that a sample can be selected to represent environmental variations over time.

[0079] The historical data set is separated to provide both training and testing data for the regression model which will predict the most efficient drying configuration.

[0080] In the described example, we will describe the training of a categorical boost (CatBoost) regression model using the training data obtained from the historical data set (after the removal of data which does not concur with the stable production parameters has taken place). However, this is for the sake of illustration only and other regression models (both linear and non-linear regression approaches) can be adopted. A CatBoost regression model is particularly effective for handling data sets with categorical features, i.e. features which take on levels or values. The historical data related to the plasterboard drying process does contain categorical features such as, for example, quality levels of the produced plasterboard, moisture ranges and production dates.

[0081] In a step S214, the training of a CatBoost regression model is initialised. The number of trees in this example is initialised at 300. The learning rate in this example is initialised at 0.2. The bagging temperature in this example is initialised to 1, i.e. the weights are sampled from an exponential distribution. The loss function in this example is the root mean square error and the metric in this example is the R2 metric, i.e. the coefficient of determination which quantifies the variation in the energy consumption which is predictable from the respective parameters in the training data set. The sub-sampling in this example may be set between 0.05 to 1 depending on how much of the fraction of the dataset is to be included in the construction of each tree. The feature sampling (by level) in this example may be set between 0.05 and 1 depending on the best split for each node during the tree building process. The minimum data in this example in each leaf may be set between 1 and 100 dependent on the minimum number of samples required to create a leaf, the selection of this value controls the split creation process when the feature sampling takes place. As overfitting is not of interest here, higher values of the minimum data, i.e. between 50 and 100 are selected but in other examples lower or higher values may be selected.

[0082] In a step S216, a grid search is then initialised to optimise the depth parameter of the CatBoost regression model. Random search, Bayesian optimisation, gradient-based optimisation, evolution-based optimisation and population-based optimisation may alternatively be used to optimise the depth parameter.

[0083] In a step S218, the resulting model may then be cross-validated using the group K-fold method based on the R2 metric with five splits in the testing data set. The target variable used in the optimisation of the depth parameter is the selected one of electrical, gas or water consumption. Other cross-validation methods may also be used.

[0084] Steps S216 and S218 enables the optimal depth parameter to be selected whilst preventing overfitting. This will ensure that the resulting regression model will not provide accurate predictions for the test data set but not for new data presented to the model.

[0085] The model with the best performance according to the R2 metric is then chosen as the candidate trained regression model. This is step S220.

[0086] In a step S222, the optimised model is then evaluated on the testing data set, i.e. the data which was not used in the training process. This is conducted using the regression testing module 104 which receives the parameters of the optimised model, including the optimised depth parameter. The test data set may be retrieved from either of the data store 112 or the regression training module 102. This can be used to determine how the parameters impact the electrical energy consumption, as detailed below with respect to Figure 3.

[0087] That is to say, the production data which was not used to train the optimised model in steps S200 to S220 is then used as the testing data set.

[0088] Any one or more of steps S206 to S222 may be executed as background steps. These steps are executed as background processes, i.e. away from the user interface provided to the user who is configuring the drying environment. That is to say, the prediction of the configuration which optimises the selected one of water, gas or electricity consumption occurs with minimal user input in they occur as background processes. This means the optimisation of the drying environment as shown herein can be carried out using a regression model but without needing data-science or machine learning expertise from the user.

[0089] That is to say, any one or more of steps S206 to S222 may be automated, i.e. without further user input.

[0090] We now illustrate, using Figure 3, how the system may utilise the model optimised using steps S200 to S222 to enable the user to configure the drying environment.

[0091] In a step S300, the configuration generation module 106 determines from the output from step S222 how each of the parameters of the historical data set impacts the target variable, i.e. the electricity, gas or water consumption. This is determined by calculating the correlation between the respective parameter of the drying environment and the respective target variable, e,g, the correlation between electricity consumption and one of the air pressure values for the drying environment. This may also be run as a background process without user input.

[0092] This enables SHAP (Shapley Additive Explanations) values to be determined. This is step S302. These show a parameters impact on a change in the output of a model. A description of the determination of SHAP values is provided in "A Unified Approach to Interpreting Model Predictions" Lundberg and Lee, 31st Conference on Neural Information Processing Systems (NIPS 2017),

[0093] In determining the SHAP values, the output from the model is decomposed to determine the contribution of each parameter to the model outcome. This enables the contribution of each parameter to be determined. That is to say, the parameters which are most important for minimising the target variable are identified. The determination of SHAP values is model agnostic, so even if a different regression model is selected, the determination of SHAP values can still be used to identify the parameters most important for minimising the target variable, i.e. the minimisation of electricity consumption.

[0094] Such a set of values are illustrated in Figure 4 using a SHAP Summary Plot which offers a comprehensive view of the most important features of the model. A SHAP Summary Plot can be generated using the Python routine "model_shap_interpret". A SHAP Summary Plot enables us to interpret the model output of the trained model. Each value in the test data set is represented on the SHAP Summary Plot by a coloured dot. For each parameter, a mean average is taken over all of the values of that parameter and values below the mean are classified as low (and given a blue dot on the SHAP Summary Plot) and values above the mean are classified as high (and given a red dot on the SHAP Summary Plot). The colours red and blue are provided as examples only and other ways of distinguishing between values of the parameter may also be used, e.g. different outlines of the dot or different colours. Alternatively or additionally, the parameters values may be binned and the SHAP values used in the SHAP summary plot may be an average of the SHAP value provided by the values in each bin.

[0095] Each value of each parameter is assigned a SHAP value on the horizontal axes. The SHAP value indicates the impact of the value of that parameter on the target variable, i.e. the electricity consumption. A higher SHAP value indicates that the corresponding parameter value predicts a higher electricity consumption and a lower SHAP value indicates that the corresponding parameter value predicts a lower electricity consumption. As set out below the configuration generation module 106 will select the lower SHAP values as being part of an optimised configuration as reduced electrical energy consumption is our target for optimising our drying environment.

[0096] The vertical list on the left of Figure 4 lists the parameters in descending order with the most impactful feature first. The SHAP Summary Plot illustrated in Figure 4 shows the dryer_Zone3_exhaust_air_fan_mean parameter is the most impactful parameter as that is the parameter with the highest variance in its respective SHAP values. It can also be seen that "dryer_Zone3_air_temperature_middle_mean" is the least impactful as it has the lowest variance.

[0097] In summary, the calculation of the SHAP values based on the output and predictions from the trained model provide us with predictions about which features are the most significant with respect to electricity consumption during the drying stage. In this example, the training and testing of the trained model provides us with the conclusion that the most important parameter for determining electricity consumption is dryer_Zone3_exhaust_air_fan_mean.

[0098] It should be understood that individual drying environments will return different parameters as being the most important. This will particularly be the case if hardware components are restricted to specific ranges of parameters.

[0099] Additionally, external temperatures and external humidities will vary from environment to environment and this will impact the temperature of air which is taken into the zones through the air inlets. The provision of the data information, which comprised fields related to the external temperature, meant that this information would be built into the optimised configuration so that adjustments can be made even if the historical production data comprised other data which was taken from an environment which was of a higher or lower temperature.

[0100] The determination of SHAP values may also be run as a background process without user input. Alternatively or additionally, a variable importance plot may also be used to show how important a parameter of the historical data is when considering energy consumption.

[0101] The gains which can be obtained by using specific parameters to configure the drying environment and the variance on those parameters can also be predicted based on the model optimised using steps S200 to S222. This means that it is possible to predict which parameters can be used to configure the drying environment to optimise a selected one of water, gas or electricity consumption. We will discuss this in more detail below.

[0102] The SHAP values may be displayed in a user interface using the user interface module 108. For example, the SHAP Summary Plot illustrated in Figure 4 may be displayed using the user interface module.

[0103] The estimated values of the parameters to aim for may then be displayed using the user interface provided via the user interface module 108 as a suggested configuration. This is step S304. This is shown in Figure 5. The determination of these values is described below.

[0104] The SHAP values and the output from the testing of the model are then provided to the configuration generation module 106. The configuration generation module 106 is configured to determine the configuration which will result in reduced energy consumption using the SHAP values and the testing data. In other words, the configuration generation module 106 will use the SHAP values and the testing data to identify the parameters which will be part of an optimised configuration.

[0105] The parameter values which contribute negative SHAP values are traversed by the configuration generation module 106. This is because these values are responsible for lower electrical energy consumption. That is to say, the configuration generation module 106 identifies those parameters where there is a positive correlation between a parameter and low electrical energy consumption, even a weak positive correlation (i.e. where the correlation value is a positive value but less than 0.25) which can be seen where values are clustered together around a confined range of negative SHAP values.

[0106] Using Figure 4 as a reference and as an illustrative example, parameters which could be said to have a positive correlation are identified by the configuration generation module 106 as being dryer_Zone3_exhaust_air_fan_mean, dryer_Zone3_fresh_air_damper_mean, dryer_Zone1_air_humidity_mean and dryer_Zone3_air_temperature_middle_mean, dryer_prezone_exhaust_fan_speed_mean, and dryer_zone2_outlet_pressure_mean . The configuration generation module 106 can then identify these parameters as the important parameters for determining low electrical energy consumption. In summary, the configuration generation module 106 identifies the parameters which have a positive correlation with low electrical energy consumption.

[0107] For the sake of clarity, these parameters have the characteristic that lower values (i.e. blue values) result in lower electrical energy consumption (i.e. low SHAP values). In other examples, more or less than variables can be identified as having a positive correlation which would contribute to reduced energy consumption.

[0108] In some drying environments, user input or a dryer configuration preset may indicate that one or more parameters such as, for example, dryer_Zone2_air_humidity_mean is fixed and this may mean that this parameter is discarded. In other environments, there may be more drying zones and the number of parameters identified as positively correlated may be increased as the amount of configuration needed may be increased.

[0109] Similarly, the configuration generation module 106 is configured to identify parameters which are negatively correlated relative to low electrical energy consumption. Using Figure 4 as an illustrative reference, dryer_Zone3_air_humidity_mean, mixer_water_gauge and dryer_Zone2_fresh_air_damper_mean can be identified as parameters which are negatively correlated relative to low electrical energy consumption (based on their SHAP values). For the sake of clarity, these parameters have the characteristic that higher values (i.e. red values) result in lower electrical energy consumption (i.e. low SHAP values). In other examples, more or less parameters can be identified. In some drying environments, user input or a dryer configuration preset may indicate that mixer_water_gauge is fixed and this may mean that this parameter is discarded. In other drying environments, other parameters could be fixed. This may well be specified in step S204 or may well be specified as a background process by a setup file associated with the specific drying environment. This means that this parameter could be discarded from the method as a parameter which cannot be changed.

[0110] It should be made clear that the parameters discussed above apply to this example. Applying the method set out above to different production data with different stable production parameters may well return other conclusions which can be applied to a different drying environment. Additionally, if the quality parameters are altered, the output from the method described above could well be different.

[0111] In short, the method describes how an optimised configuration of a drying environment can be identified using historical production data with stable production parameters which are identified based on requirements of the output plasterboard.

[0112] The configuration generation module 106 is then configured to determine quantiles of the respective values corresponding to each parameter using the historic data. These are shown in Figures 5a and 5b. This enables the quantiles of the parameters to be determined to enable values to be identified which correspond to optimal performance.

[0113] Using a user interface module 108, a user interface can be provided which displays quantile ranges for each parameter identified as positively or negatively correlated by the configuration generation module 106. The user interface module 108 may highlight the 25 th< quantile for positively correlated features and the 75 th< quantile for negatively correlated features using a corresponding indicia such as outline box 502 in a user interface 500. In Figures 5a and 5b such a user interface is visualised. For clarity, Q 25, i enumerates the 25 th< quantile of variable i and Q 75,i enumerates the 75 th< quantile of variable i.

[0114] In other examples, other quantiles may be highlighted by the configuration generation module 106. These quantiles may be highlighted using indicia in the form of a line or even in the form of colours, words or other alphanumerical sequences characters. Alternatively or additionally, any values may be highlighted using indicia in the form of a line or using colours, words or other alphanumerical sequences characters. For example, if the configuration generation module 106 has determined a zero value for a parameter then it may also be highlighted.

[0115] The user interface module 108 may provide the highlighted portions, i.e. the 25 th< and 75 th< quantiles as selectable hyperlinks which, on selection by the user, can provide input to the system to indicate that the corresponding parameter set is to be selected for configuration of the dryer environment. This is because the configuration generation module 106 has identified that configuration as optimal for reducing electrical energy consumption.

[0116] The configuration generation module 106 may also use the quantile values and the test data to determine the electrical energy consumption associated with the quantile value. This can then be displayed in user interface 500 alongside the data shown in Figure 5. The configuration generation module 106 may also then calculate the gain or loss in electrical energy consumption as a result of choosing one quantile over another. Figure 6 shows an example of such a display where, for a selection of the parameters, the 25 th< quantile is compared to the 75 th< quantile and the respective gain or loss in electrical energy consumption is displayed as a percentage gain or loss when one value is selected compared to another. In some embodiments, all of the parameters may be compared to identify the gain obtained by selecting one parameter over the other. That is to say, the configuration generation module 106 shows that by selecting the Q25 value, i.e. 25 th< quantile, of the dryer_Zone3_exhaust_air_fan_mean it is possible to realise an electrical energy cost saving of 3.8329 percent relative to the Q75 value, i.e. 75 th< quantile of the same parameter.

[0117] This enables the operative to carefully tune the drying environment if necessary and realise a trade-off between parameters which both increase and decrease electrical energy consumption.

[0118] The user interface module 108 then, in a step S306, provides a prompt which enables the user to select the configuration which is highlighted by the user interface module 108 based on the predictions provided by the model optimised in steps S200 to S222. That is to say, the configuration generation module 106 can efficiently enable the user of the system 100 to configure a drying environment using the optimised parameters illustrated in Figure 5. Additionally or alternatively, the configuration generation module 106 enables the user of the system 100 to carefully tune the drying environment to realise electrical energy savings.

[0119] The effect of this is that historical data can be used to configure a drying environment to consume minimal electrical energy whilst still realising desired quality levels.

[0120] That is to say, the optimal values of the parameters identified as having most impact on the consumption of the selected one of electricity, water or gas are provided to the user and they are the provided with a user interface which enables them to select that configuration of the drying environment.

[0121] The selected values are then used to generate a configuration file in a step S308 in that the selected values are identified as being the optimal values for a drying environment. The configuration file forms an output data set which can be used to configure the drying environment for the drying of plasterboard. The configuration file includes default parameters for a drying environment in addition to the optimised parameters of the most correlated values. The configuration file may include user selected parameters for the drying environment in addition to the optimised parameters of the most correlated values which have been identified using the processes which are identified in Figures 2 and 3.

[0122] The configuration file is then transmitted to the drying interface 110 for the drying environment which is being configured by the user. This is step S310. This is then used by the operating system at the drying environment to configure the drying environment so that it can be used to provide plasterboard of the required quality whilst optimising energy consumption. That is to say, in the discussed example, the values identified from the historical production data are used to configure the drying environment. This means that the components (e.g. dryer temperatures, dampers etc) are configured in accordance with the identified optimal parameters.

[0123] Alternatively or additionally, steps S200 to S222 and steps S300 to S310 may be run for two or more separate and independent target variables to determine an optimal configuration which requires the optimisation of two of electricity, water or gas. That is to say, the methods may be run in parallel if, for example, the user was interested in configuring a drying environment where the ventilation fans were run using electricity by the heating was run using gas.

[0124] The drying environment may comprise, for example, a longitudinal or transversal dryer. An example section of a longitudinal dryer is illustrated in Figure 7 and a full, 3-zone longitudinal dryer is shown in Figure 7a. An example transversal dryer is illustrated in Figure 8 where a cross-section of the dryer is also shown.

[0125] A longitudinal dryer may comprise a pre-drying zone (the pre-zone) and three drying zones as illustrated in Figure 7a. Each drying zone may comprise a return plenum, an inlet plenum, an exhaust outlet, a fresh air inlet, a combustion air inlet, a natural gas inlet and a recirculation fan. Plasterboard may be passed through each of the zones where the natural gas is used with the combustion air provided through the combustion air inlet to heat up the fresh air provided through the fresh air inlet in order to generate a counter-current of hot air which is circulated by the recirculation fan into a drying chamber. The counter-current of hot air passes through the drying chamber in the opposite direction to the path along which the drying plasterboard is being moved. The counter-current of hot hair causes the plasterboard being moved through the drying chamber to dry to a level which is specified by stable production parameters for the manufacture of that consignment of plasterboard.

[0126] A transversal dryer may comprise a plurality of zones as illustrated in Figure 8 where a transversal dryer with N zones is illustrated schematically. N is typically between 20 and 30. Layers of plasterboard are input into dryer zone 1 and they are then carried through the respective zones by any suitable means Each section may be between 2 and 4m in length.

[0127] The layers of plasterboard (boards) are passed through each of the zones during the drying process of the consignment of plasterboard Heated air generated using any suitable means e.g. a burner is distributed over the boards by a blowing manifold and a return manifold then directs the air upwards. Part of the current of air exits through an exhaust and part of the current of air is recycled by the burner where it is reheated and again directed toward the blowing manifold.

[0128] The configuration file transmitted in step S310 is used to configure the different components of each dryer zone (including the pre-zone). Temperatures, fan speeds, line speeds and the like are each specified by the parameters provided in the configuration file. That is to say, the trained model is used to configure the drying environment in a way which optimises the selected target value. Where multiple, e.g. 20-30 dryer zones are used, the configuration file may specify each zone differently.

[0129] In an example, if the selected target value is electricity, each dryer zone is configured to minimise the consumption of electricity whilst providing plasterboard at the desired moisture level, i.e. at the desired moisture level which is defined in step S204. The drying environment is operated using a control system which is configured to execute the instructions provided by the operating system. This is illustrated in Figure 9. The configuration file generated in step S308 and then transmitted to the drying environment interface in step S310 is then received by the drying environment interface 110 at the drying environment 122. Each drying environment 122 is controlled by a control system 120. The control system 120 is a system which controls the components of the drying environment e.g. the frequency of the recirculation fan or the amount of air provided through the combustion air inlet.

[0130] The drying environment interface 110 feeds the configuration file to the operating system which instructs the control system 120 using the parameters inside the configuration file. The operating system then uses the parameters inside the configuration file to configure the drying environment using the control system 120.

[0131] That is to say, the configuration file generated as a result of steps S200 to S222 and then steps S300 to S310 is used to control the drying environment.

[0132] In the example set out above, this means that, based on the SHAP values and the predictions generated using the regression model the dryer_zone3_exhaust_air_fan_mean is set at, for example, 75 to optimise energy usage whilst maintaining the quality of plasterboard (e.g. s characterised by the average moisture level) required by the stable production parameters which are identified in step S204. This is because the training and testing of the model identifies a correlation between the dryer_zone3_exhaust_air_fan_mean and the energy consumption.

[0133] This optimised configuration is generated using a regression model trained using historical data associated with previous plasterboard production which met the same stable production parameters. The drying environment is configured with minimal user input in that the user specifies the stable production parameters and the energy consumption parameter e.g. electricity but the training and testing of the regression model is carried out as a background process and without the need for the user to possess competence in regression models and data science.

[0134] Given different stable production parameters, the model may be used to predict, for example that drying using a transversal dryer optimises energy usage if, the burner generates a hot air flow at a specific temperature, for instance. This can then be used to configure a drying environment comprising such a burner to generate hot air at the specific temperature. The inclusion of the stable production parameters means that the required quality of plasterboard is achieved whilst the energy consumption is optimised.

[0135] It should be noted that the above-mentioned aspects and embodiments illustrate rather than limit the disclosure, and that those skilled in the art will be capable of designing many alternative embodiments without departing from the scope of the disclosure as defined by the appended claims. In the claims, any reference signs placed in parentheses shall not be construed as limiting the claims. The word "comprising" and "comprises", and the like, does not exclude the presence of elements or steps other than those listed in any claim or the specification as a whole. In the present specification, "comprises" means "includes or consists of" and "comprising" means "including or consisting of". The singular reference of an element does not exclude the plural reference of such elements and vice-versa. The disclosure may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In a device claim enumerating several means, several of these means may be embodied by one and the same item of hardware. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

Claims

1. A computer-implemented method of generating configuration data which can be used to configure a drying environment for the drying of plasterboard, the method implemented by a processing resource, the method comprising: - obtaining a first set of historical data associated with the drying of plasterboard, wherein the historical data comprises data associated with energy consumption measurements for the drying of plasterboard; - using the historical data to identify parameters which increase or decrease energy consumption in the process of drying of plasterboard, wherein using the historical data to identify parameters which increase or decrease energy consumption comprises applying a trained model to the historical data to identify the parameters which correspond with an increase or decrease in energy consumption; - providing an output data set using the trained model, the output data set identifying parameters which minimise energy consumption in the drying of plasterboard, wherein the trained model is used to predict the parameters which will result in the minimisation of energy consumption during the drying of plasterboard.

2. A method according to Claim 1, wherein the method further comprises providing a user interface displaying the parameters which minimise energy consumption in the drying of plasterboard.

3. A method according to Claim 1 or Claim 2, wherein the method further comprises generating a configuration file for the drying environment based on the parameters which minimise energy consumption.

4. A method according to Claim 3, wherein the configuration file comprises parameters identified by the trained model as being associated with minimisation of energy consumption during the drying of plasterboard.

5. A method according to any preceding claim wherein the method further comprises transmitting the configuration file to a drying environment.

6. A method according to Claim 2, wherein the configuration file is generated responsive to user input requesting the configuration file.

7. A method according to any one of Claims 4 to 6, wherein the configuration file is generated responsive to user input selecting parameters for the drying environment using the user interface.

8. A method according to Claim 7 wherein the parameters for the drying environment are identified using the trained model.

9. A method according to Claim 7 or Claim 8, wherein the parameters are identified using indicia which is used to identify parameters associated with an optimised configuration predicted by the model.

10. A method according to any preceding claim, wherein the trained model identifies the gain or loss in energy consumption associated with a parameter.

11. A method according to any one of Claims 1 to 10 wherein the user interface identifies positive and negative correlations associated with parameters.

12. A method according to any preceding claim, wherein the identification of parameters associated with a configuration is executed as a background process.

13. A system configured to implement the method of any one of Claims 1 to 11.

14. A computer-program product which, when executed on a processing medium, configures the processing medium to implement the steps of any one of Claims 1 to 11.

15. A non-transitory storage medium configured to store instructions which, when executed by suitably configured hardware, provides instructions to a processing medium to implement the steps of any one of Claims 1 to 11.

Images