Computer-implemented method for estimating a parameter of concrete in a drum of a concrete mixing vehicle and related system

Abstract

Computer-implemented method (30) for calculating an estimate of a parameter characterising concrete (e.g. slump) in a drum (101) of a concrete mixing vehicle (100), comprising the steps of: - receiving (31) a drum speed value indicative of a rotational speed of the drum; - receiving (31) a torque value indicative of a torque required to turn the drum; - providing (32) the drum speed value and torque value as input measurements to a relationship estimator Bayesian filter; - calculating (33) at least one vehicle parameter characterising a relationship between drum speed and torque using the relationship estimator Bayesian filter, wherein the relationship estimator Bayesian filter is configured such that the at least one vehicle parameter comprises a hidden state variable of the relationship estimator Bayesian filter; and - calculating (34) an estimated parameter value by providing the at least one vehicle parameter as an input to a parameter calculation model.

Description

[0001]SLUMP PREDICTION SYSTEM FIELD OF THE INVENTION The invention relates to systems and method for estimating slump in concrete delivery vehicles. BACKGROUND TO THE INVENTION Concrete is a key material for modern construction techniques, where it is highly valued for its strength and versatility. Concrete can be prepared in liquid form and then poured, whereupon it sets and hardens into a solid. Liquid concrete is often produced offsite at batching facilities and loaded into a concrete mixing vehicle for delivery to a construction site. Concrete mixing vehicles are typically trucks incorporating a mixing drum for carrying a load of concrete. The drum is spun by a motor during operation, mixing the concrete and preventing it from setting during delivery. The mechanical properties of concrete can vary significantly between concrete batches. A widely used measure for characterising the consistence or workability of freshly mixed concrete is termed slump. It is a manual test, defined by the displacement of the concrete when a standardized cone-shaped testing mould is removed, with the difference in height between the top of the testing mould and the displaced concrete serving as the slump value. This metric provides valuable insights into the quality and performance of the concrete mix. A higher slump value indicates a more fluid mix, which may be desirable for applications such as pouring into intricate forms or congested reinforcement. Conversely, a lower slump value signifies a stiffer mix, suitable for vertical placements or structural elements requiring greater strength and durability. By accurately measuring slump, construction professionals can ensure the produced concrete meets the specific requirements of each project, optimising workability, strength, and durability while minimising the risk of defects or structural failures. In addition to the conventional slump test, consistence can also be measured using a slump-flow test or a flow-table test or other equivalent tests. The consistence measurement is a proxy for certain rheological properties of the concrete. Methods for measuring the consistence and rheological properties of fresh concrete on- board concrete mixing vehicles are also known. Such methods are often advantageous, compared to having to stop the vehicle to perform a test enroute, or waiting until the vehicle gets to its destination. However, existing methods usually require the drum to spin at a pre-determined constant speed or within a speed range. This can be impractical because the drum motor is generally powered by the vehicle’s engine, meaning that the speed of the drum is dependent on the speed of the engine. In some cases, a control system may be fitted to a concrete mixing vehicle to fix or control the speed independently of the engine. However, such control systems add significant mechanical complexity, weight, and cost, and may increase the vehicle’s power consumption. Accordingly, there exists a need for improved systems and methods for monitoring slump in concrete delivery vehicles. SUMMARY OF THE INVENTION According to a first example, there comprises a computer-implemented method for calculating an estimate of a parameter characterising concrete in a drum of a concrete mixing vehicle, comprising the steps of: receiving a drum speed value indicative of a rotational speed of the drum; receiving a torque value indicative of a torque required to turn the drum; providing the drum speed value and torque value as input measurements to a relationship estimator Bayesian filter; calculating at least one vehicle parameter characterising a relationship between drum speed and torque using the relationship estimator Bayesian filter, wherein the relationship estimator Bayesian filter is configured such that the at least one vehicle parameter comprises a hidden state variable of the relationship estimator Bayesian filter; and, calculating an estimated parameter value by providing the at least one vehicle parameter as an input to a parameter calculation model. This method provides an improved means for monitoring parameters characterising concrete (such as slump) in concrete mixing vehicles, which is more robust, adaptable, and accurate compared to existing approaches. It applies a Bayesian approach, whereby the vehicle parameters of the drum speed vs torque relationship have an associated uncertainty (and during an initial period, may have a high associated uncertainty). The drum speed and torque may then be sampled periodically or an event driven manner (e.g. every drum rotation) and the measured values can be used to update the current estimate of the vehicle parameters, in a way that accounts for the uncertainty or variability of the measurements. Advantageously over time, the uncertainty associated with the vehicle parameters may decrease and their reliability may increase. Rather than directly inferring a parameter estimate value from an instantaneous drum speed and torque value, the method uses a Bayesian filter to calculate an estimate of one or more vehicle parameters. It is the vehicle parameters which are used to calculate the parameter estimate using the parameter calculation model. A Bayesian filter (also known as a Bayes filter) encompasses any filter based on Bayes' theorem, which describes how to update the probability of a hypothesis based on new evidence. Bayesian filters may be recursive, meaning that at each calculation interval, the filter uses its own previous output from the preceding calculation interval. As will be explained in further detail below, the choice of vehicle parameters to be estimated by the relationship estimator Bayesian filter may be selected for their insensitivity to fluctuations in drum speed and torque. In other words, the relationship modelled by the relationship estimator Bayesian filter is chosen to accommodate a range of drum speeds and torques. This enables the calculation of accurate parameter estimates continuously at consistent calculation intervals, even if the drum speed and torque are constantly changing, and / or if the measurements are noisy. In some examples, two vehicle parameters may be used. These may for example represent the gradient and offset, respectively, of a line of best fit between torque and speed. Furthermore, the vehicle parameters of the relationship between drum speed and torque hold more information about the rheology of the concrete than a single slump test or the pressure alone (e.g. the plastic viscosity). A direct mapping of torque (or pressure in the case where the drum drive motor is hydraulic) at a given speed provides a limited description of the rheological behaviour of the freshly mixed concrete. For fluids that behave like fresh concrete, a more complete description of workability and / or consistence would comprise the relationship between speed and torque in terms of at least two parameters (e.g. the intercept and slope of a linear relationship). In some examples, the vehicle parameters may represent a straight-line linear relationship between drum speed and torque. In other examples, rather than using modelling the relationship as a straight line, a different linear or non-linear relationship could be modelled. For example, the best-fit relationship could be modelled as a square or cubic equation, governed by a set of parameters. The relationship between drum speed and torque may be modelled off an established rheological relationship, such as the Bingham plastic model, or from a fluid model such as the Herschel–Bulkley model. The drum speed value and torque value may directly represent the drum speed and torque. Alternatively, the drum speed value and torque value may indirectly represent the drum speed and torque, e.g. by representing values which are correlated to the drum speed or torque. For example, the torque value may be the pressure differential across a hydraulic motor used to drive the drum. A hidden state variable of a Bayesian filter refers to a variable that represents the unobservable or latent aspects of the system being modelled. The hidden states are estimated by the Bayesian filter estimate based on the available measurements. In this method, the vehicle parameters are hidden state estimates because they represent a relationship between torque and drum speed, rather than a quantity which can be physically measured. Optionally, the relationship estimator Bayesian filter comprises a Kalman filter. Optionally, the relationship estimator Bayesian filter comprises a non-linear model for modelling the relationship between drum speed and torque. Kalman filters are a sub-type of Bayesian filters, which can advantageously be easily configured for state estimation. In some examples, the Kalman filter may be a “standard” Kalman filter, which uses a linear model and assumes that uncertainty and noise can be modelled as Gaussian distributions. In other examples, a non-linear model may be used, and the Kalman filter may be an extended Kalman Filter. Variations of the standard Kalman filters may be used, such as an extended Kalman filter, unscented Kalman filter, and other examples known in the art. Optionally, the method further comprises the steps of: receiving batch information for the concrete in the drum of the concrete mixing vehicle, and, selecting (or pre- selecting), based on the batch information, the parameter calculation model from a set of stored parameter calculation models. Advantageously, this allows a parameter calculation model to be selected which is most suited to the particular characteristics of the concrete in the drum. Optionally, the inputs to the parameter calculation model additionally comprise one or more of: the volume of concrete in the drum; and the temperature of the concrete in the drum. The volume information may be received from an external source, such as a customer docket. Alternatively, the volume of concrete in the drum may be determined by a sensor on the concrete mixing vehicle, such as a weight sensor. The temperature of concrete in the drum may be determined by a direct sensor, i.e. a sensor which reads the temperature of the concrete by direct or electromagnetic sensing. The temperature of the concrete in the drum may also be determined using an indirect sensor, i.e. a sensor which directly measures the temperature of another component, such as the skin of the mixing drum, and from this measurement the temperature of the concrete can be inferred. Optionally, the drum speed value represents the average drum speed over a time period and the torque value represents the average drum torque over said time period. Averaging the drum speed and torque values over a time period reduces the effect of anomalous measurements and / or random noise. The amount of data to be processed may also reduced, because a parameter calculation may be calculated once every averaging time period rather than at the measurement sampling rate. Optionally, the method may further comprise the steps of: generating a standard deviation value representing one or more of the drum speed standard deviation over said time period, or the torque standard deviation over said time period, calculating, from the standard deviation value, a measurement noise value, and providing the measurement noise value as an input to the relationship estimator Bayesian filter. The standard deviation associated with a mean drum speed / torque value has been found to provide valuable information. A low standard deviation indicates that the measurements over the time period were similar values, indicating that either the sensor noise is low or the system itself is in a more stable state. Conversely, a high standard deviation indicates either high measurement noise or more fluctuation in the system. Hence, a low standard deviation may indicate low measurement noise and vice versa. By varying the process noise of the relationship estimator Bayesian filter, the accuracy and responsiveness of the filter can be adjusted. It will be understood that variance may be calculated and used in the place of standard deviation. Variance is the square of standard deviation, so it is mathematically trivial to substitute one for the other. Optionally, the drum speed value and torque value are received at a first periodic interval; the received drum speed value and torque value are sampled at a second periodic interval which is slower than the first periodic interval, to generate a sampled drum speed value and torque value; and the one or more vehicle parameter and the estimated parameter value are calculated at the second periodic interval using the sampled drum speed value and torque value. For example, the drum speed and torque value may be sampled at 0.01 second intervals, and then the one or more vehicle parameters may be calculated at 10 second intervals. This enables sensor sampling and initial processing (which are less computationally intensive) to be carried out at high frequency, while the following estimation steps (which are more computationally intensive) are carried out at a lower frequency, thereby reducing overall computational requirements. Data transfer is also reduced, for cases where subsequent processing is on another device. Optionally, the method further comprises the steps of: receiving a rotation measurement from a rotation sensor and a torque measurement from a torque sensor; and, processing the received rotation measurement and torque measurement to respectively generate the drum speed value and torque value. The torque sensor may comprise a sensor system in the sense that the actual measurements may be performed by two or more sub-sensors. The torque measurement may be an indirect measurement, that is, a measurement or measurements of another value which is indicative of torque and / or from which the torque can be calculated. For example, the drum may be driven by a hydraulic motor, and the torque measurement may comprise one or more pressure measurements of the hydraulic pressure, (e.g. one taken upstream of the hydraulic motor and one taken downstream). The pressure measurement (or differential pressure between two measurements) can be used to calculate the torque. Alternatively the drum may be driven by other means, e.g. an electric motor. In such examples, the torque sensor may comprise an electrical sensor configured to measure the current through and / or voltage across the electric motor, and these values then comprise the torque measurements which are used to calculate the torque value. Similarly, the rotation sensor may perform only indirect measurement of rotation. For example, the rotation sensor may comprise an accelerometer mounted to a wall of the drum and configured to rotate with the drum. The rotation measurement in this example would comprise one or more acceleration measurements, which could be processed to calculate a drum speed value by means of e.g. an integration calculation. In other examples, the rotation sensor may sense drum speed or rotation directly. For example, the rotation sensor may comprise a speed ring sensor. A speed ring sensor comprises a slotted or marked ring attached around a circumference of the drum and configured to rotate with the drum, and a stationary sensor configured to count the slots or markings as the drum rotates past. By identifying the marking / slot or measuring the rate at which markings / slots pass by, a rotation angle and drum speed value can be calculated. Optionally, processing the received rotation measurement comprises the steps of: providing the rotation measurement to a drum estimator Bayesian filter; calculating the drum speed value using the drum estimator Bayesian filter, wherein the drum estimator Bayesian filter is configured such that the drum speed value comprises a hidden state variable of the drum estimator Bayesian filter. Using a drum estimator Bayesian filter to estimate the drum speed value may advantageously improve accuracy and robustness compared to e.g. a simple direct integration of the accelerometer readings. This in turn improves the accuracy and robustness of the parameter estimate. This is because, as previously discussed, a Bayesian filter accounts for the uncertainty associated with incoming measurements and its own internal hidden state variables. For example, the drum estimator Bayesian filter speed may be able to accommodate noise in the incoming accelerometer measurement, as well as filtering out other sources of acceleration (such as jolting of the concrete mixing vehicle over bumps or lateral acceleration around corners). Optionally, the drum estimator Bayesian filter comprises a Kalman filter. Optionally, processing the received torque measurement comprises the steps of: providing the received torque measurement as an input to a sliding window averaging filter, and recording as the torque value an output of said filter. Optionally, the method further comprises the step of recording the torque value standard deviation as an output of the sliding window averaging filter. That is, the sliding window averaging filter is also configured to output the standard deviation of the torque value. The operation of a sliding window filter is to take a window of a dataset and calculate the value of a parameter across the window. For example, a sliding window may take in a series of datapoints as an input and calculate the mean of the data over the window as its output. This process is repeated for every new data point and the sliding window maintains a subset (n-1) of the historical values to use in each new calculation. The sliding window averaging advantageously smooths out anomalies and sensor spikes in the torque value. Further advantageously, sliding window filters may be implemented without significant computational resources. Optionally, the method further comprises the steps of: transmitting, from a local transmitter, the drum speed value and torque values; and receiving, at a remote receiver, the drum speed and torque values. In other words, the method may be carried out in a distributed manner. This may enable computational tasks to be allocated to different locations depending on the computational resources required. For example, the comparatively less computationally intensive steps of calculating a drum speed value and torque value from sensor measurements may be carried out by a local processor. The drum speed value and torque value are transmitted to a remote location where a remote processor completes the remaining, more computationally intensive steps of calculating a parameter estimate. “Local” as used herein may refer to a system located or process being carried out onboard a concrete mixing vehicle, whereas “remote” may refer to a system located or process being carried out at any other location. A “remote location” may be a distributed location, in the sense that different parts of the method are carried out in geographically separate areas. For example, the remote receiver may be located at a first remote location, and then communicate or relay the received values to a processing unit located at a second remote location. In some examples, the method may be performed by a distributed cloud processing unit (i.e., the method may be performed in the Cloud). Optionally, the method of any preceding claim further comprises the steps of: determining, from the drum speed value or drum torque value, whether the concrete mixing vehicle is in a discharge mode, and if the vehicle is in a discharge mode, switching the relationship estimator Bayesian filter from a normal mode to an alternate mode. Discharge refers to the process of expelling concrete from the mixing drum. Typically, this is done by rotating the mixing drum in the opposite direction to the normal charging mode. Thus, it may be determined that discharge is (or is about to be) occurring when the drum speed or torque value change in sign (e.g. go from a positive value to a negative value). In the alternate mode, the relationship estimator Bayesian filter may be configured to use a different internal model, or to assign a different weight to the drum speed value or torque value, to account for the different characteristics of the concrete and concrete mixing vehicle when it is discharging. Optionally, the relationship estimator Bayesian filter comprises a weighting value for weighting the influence of the drum speed value or torque value as input measurements, and the weighting value is lower in the alternate mode compared to the normal mode. During discharge, the rotation sensor and / or pressure sensor may become less reliable. For example, the torque value may begin decreasing during discharge as the volume of concrete in the drum decreases and the drum becomes easier to turn. Lower-weighting the drum speed or torque values during discharge (i.e. in the alternate mode) ensures that the estimates generated by the relationship estimator Bayesian filter are not pulled away from their true values due to incorrect or unreliable measurements. In some examples, both the drum speed value and torque value are lower-weighted in the alternate mode. In other examples only the drum speed value or torque value is lower-weighted. Optionally, the weighting value is zero in the alternate mode. In other words, the drum speed and torque value are disregarded and not used as inputs to the relationship estimator Bayesian filter. In some examples of the method, the drum speed value and torque value are not even provided to the relationship estimator Bayesian filter when it is in the alternate mode. The relationship estimator Bayesian filter may continue to provide an estimate for the vehicle parameters even when the weighting value is zero, by generating the estimates as predictions from the relationship estimator Bayesian filter’s internal model. By generating this prediction, calculation of a parameter from the drum speed and torque values can be resumed following a sample pour. The method can be used for estimating a variety of parameters characterising concrete. For example, as well as slump, the method may be used to calculate an estimate of the flow table test result (similar to slump, but generally used to characterise low viscosity concrete), slump-flow, consistence, or other parameters including rheological parameters. It is increasingly recognised that traditional methods for measuring the consistence of fresh concrete, like the slump test, may be inadequate for modern concrete mixes, which require more precise quality control. A single value of slump may be insufficient to fully characterize today’s advanced mixes like self-compacting or 3D-printed concrete, which demand more sophisticated quality control to prevent segregation and ensure proper placement. Hence in some examples, the method may be used to characterise concrete using two or more rheological parameters. For example, the relationship between drum speed and torque may be modelled off the Bingham model, with the estimated rheological parameters representing yield stress (resistance to initial flow), and plastic viscosity (flow resistance once flow has commenced). Such parameters provide a more complete description of fresh concrete's flow behaviour, enabling a wider variety of concrete types to be characterised with greater accuracy and facilitating modern construction techniques such as pumping. Rheometers capable of analysing the relationship between shear stress and shear rate are known. However current state-of-the-art rheometers designed to measure these properties have limitations. Although they provide precise measurements of yield stress and plastic viscosity, their adoption in the construction industry is hindered by several drawbacks. These devices are costly, require skilled operators, and are not suited for field conditions, making them impractical for on-site quality control. Consequently, there is a pressing need for simpler, field-deployable methods that can still fully characterize the flow properties of fresh concrete, ensuring more accurate and reliable quality control for modern concrete mixtures and placing technologies. The method of the first aspect can effectively transform a standard concrete mixing vehicle into a “mobile rheometer”, representing a significant advancement in quality control for concrete. According to a second aspect, there comprises a system for calculating an estimate of a parameter of concrete in a drum of a concrete mixing vehicle, comprising a processor configured to carry out the method according to the any previous example. Optionally, the processor is a remote processor. Optionally, the system further comprises: a local processor; a local transmitter; a remote receiver; and, wherein the processor is configured to: receive a rotation measurement from a rotation sensor and a torque measurement from a torque sensor; and, process the received rotation measurement and torque measurement to respectively generate the drum speed value and torque value, and wherein the local transmitter is configured to transmit the generated drum speed value and torque value to the remote receiver, and wherein the remote receiver is configured to receive the drum speed value and torque value and communicate them to the remote processor. The local processor may also be configured to receive the rotation measurement and / or torque measurement and generate the drum speed value and torque value, as previously described (e.g. using a drum estimator Bayesian filter and a torque sliding window average filter). The local processor may also be configured to calculate the standard deviation of the drum speed and / or torque, as previously described. The local processor may be configured to sample the generated drum speed value and torque value from a first periodic interval to a second periodic interval as previously described. The sampled generated drum speed value and torque value may then be transmitted to the remote processor. According to a third aspect, there comprises a non-transitory computer-readable storage medium storing instructions to cause the processor of the second aspect to perform the method of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 shows a concrete delivery vehicle according to an example. Figure 2 shows a schematic of a processing system according to an example. Figure 3(a) and 3(b) are flowcharts illustrating methods for calculating a slump estimate according to examples. Figure 4 illustrates processing elements of a system 400 configured to calculate a slump estimate. Figure 5(a)-(b) show graphs of drum speed against torque. Figure 6 shows a graph of drum speed against torque, illustrating the calculation process of an exemplary relationship value Bayesian filter. Figure 7 is a flowchart illustrating a method for calculating a slump estimate during discharge modes, according to an example. DETAILED DESCRIPTION For convenience, the examples discussed herein will relate to calculating an estimated slump. However as previously discussed, other parameters such as consistence or rheological parameters may also be calculated. Figure 1 shows an exemplary concrete mixing vehicle 100. The vehicle 100 comprises a drum 101, which may be filled with concrete. The drum 101 is driven by a motor 102, which in this example is a hydraulic motor. The hydraulic motor 102 is itself connected to and driven by the engine 100A of the vehicle 100. The hydraulic motor 102 comprises a hydraulic line 102A, extending upstream and downstream of the motor 102. A rotation sensor 103 is configured to sense rotation of the drum 101. In this example, the rotation sensor 103 comprises an accelerometer that mounted to the exterior of the drum 101 and configured to rotate with the drum 101. A torque sensor 104 is configured to sense the torque required to turn the drum 101. In this example, the torque sensor 104 comprises a pair of pressure sensors 104A, 104B, configured to sense the hydraulic pressure in the hydraulic line 102A immediately upstream and downstream of the hydraulic motor 102. The torque sensor 104 and accelerometer 103 are connected to a local processing device 110 onboard the vehicle 100, either wirelessly or through wired connections. The local processing device 110 is configured to periodically sample the sensors and obtain measurements. The local processing device 110 may comprise one or more processors in communication with memory. The local processing device 110 is connected to a transmitter 106 (which may in some examples also function as a receiver). In some examples, the local processor 110 is configured to carry out an entire exemplary slump estimation method, as previously described. The calculated slump estimate may then be displayed to an operator via an output device, or it may be transmitted to a remote location via the local transmitter 106. In other examples, the local processing device 110 carries out only part of the steps of an exemplary slump estimation method, such as the steps in between receiving measurements from the sensors and generating the drum speed value and torque speed value, and then transmits these generated values to a remote processor via the local transmitter 106. Figure 2 illustrates a schematic of a processing system for implementing an exemplary method. In this example, the drum speed value and torque value are generated by the processing device 110 on board the concrete mixing vehicle 100 of Figure 1. The generated values are transmitted by the transmitter 106 to a remote location 200, where they are received by a receiver 206. The generated values are then passed to a processing device 210. The processing device 210 comprises one or more processors 212 in communication with memory 214. The memory 214 is an example of a computer readable storage medium. The one or more processors 212 are also in communication with one or more output devices 218. The various components of the processing device 210 may be implemented using generic means for computing known in the art. For example, the output devices 218 may comprise a monitor or display. The output devices 218 may be used to display one or more calculated slump estimates to a user, or representations thereof (such as e.g. a graph of estimated slump over time). The calculated slump estimates may also be stored in the memory 214. In addition, the processing device 210 comprise a network access device 220 such as a modem or network adaptor, to provide access to a network to obtain the generated values from the receiver 2. In some examples, the processing device 210 may be configured to transmit the calculated slump estimate to another remote location, whereupon it is communicated or displayed to an operator. This may cause the operator to take an action in response. For example, the operator may change the route of the vehicle 100 from a first delivery site to a second delivery site. An exemplary method 30 for calculating a slump estimate is now described in relation to Figure 3(a). In step 31, a drum speed and torque value are received. In step 32, the drum speed value and torque value are provided as input measurements to a relationship estimator Bayesian filter. In step 33, at least one vehicle parameter characterising a relationship between drum speed and torque are calculated using a relationship estimator Bayesian filter, wherein the relationship estimator Bayesian filter is configured such that the at least one vehicle parameter comprises a hidden state variable of the relationship estimator Bayesian filter. In step 34, an estimated slump value is calculated by providing the at least one vehicle parameter as an input to a slump calculation model. A more detailed exemplary method for calculating a slump estimate is now described in relation to Figure 3(b). In step 301, a rotation sensor configured to sense the rotation of the mixing drum of a concrete mixing vehicle is sampled. In this example, the rotation sensor is an accelerometer. In some examples, the accelerometer sensor comprises at least two accelerometer sub-units, each of which detects acceleration along a particular axis. The accelerometer sub-units are arranged such that their sensing axes are perpendicular to each other and also to the axis of rotation of the drum. Sampling the accelerometer may comprise sampling the measurements of the sub-units. The measurements may be combined into a single measurement of drum angle by identifying the direction of the gravity vector in the accelerometer readings. In step 302, the sampled accelerometer readings are provided as inputs to a drum estimator Bayesian filter, which in this example comprises a Kalman filter. The Kalman filter is configured such that the (i) estimated drum angle ; (ii) average (in this case mean) drum rotational speed ; and (iii) estimated standard deviation of drum rotational speed are state variables. These hidden state variables are outputted for each sampled input measurement. The drum estimator Bayesian filter may be a Kalman filter. It may use a constant process noise, which may be selected to represent the best balance between the estimated angle of the drum and the newly measured angle of the drum, and may be a predetermined constant, non-zero value. The process noise may be determined experimentally for best performance. The drum estimator Bayesian filter may also apply a constant measurement noise. This value may represent the level of confidence in the sampled accelerometer measurements, and may be set to a predetermined constant, non-zero value. This value may also be determined experimentally and may be based on the assumption that every measurement is afforded a preset level of reliability or confidence. In step 303, the estimated average or average drum speed and the estimated drum speed standard deviation outputs from the drum Bayesian filter are subjected to a low-pass filter. The purpose of the low-pass filter is to introduce a small time delay for the drum speed values, to assist in aligning the timing with that of the pressure sensing calculation (which will be discussed below). The low- pass filter may comprise, for example, an FIR or IIR filter. In step 304, the drum angle and the outputs of the low-pass filter are sampled. In some examples, the drum estimator Bayesian filter and the low-pass filter may operate at a first calculation interval or clock speed, but their outputs may be sampled at 304 at a second calculation interval. For example, the first calculation interval may be 0.01 seconds (averaging that a new value is calculated at 100 Hz), and the second calculation interval may be 10 seconds. In other words, only every 1000th output is sampled. In this example the drum speed value comprises the estimated average drum rotational speed , once it has been passed 303 through the low-pass filter and sampled 304. The process for generating the torque speed value begins at step 305, in which measurements from a torque sensor are sampled. In this example, the torque sensor comprises pressure sensors upstream and downstream of a hydraulic motor, and sampling the torque sensor comprises sampling both pressure sensors.In step 306, the readings from each pressure sensor are used to calculate the average, and standard deviation , for each respective pressure sensor. In thisa second sliding window filter is used to calculate these values. The output of the sliding window filter is sampled at the same sampling interval and sampling frequency as performed in step 304. In this example, the torque speed valuecomprises two values , , as the difference between the average pressures,is proportional to the torque required to turn the drum.In step 307, the calculated and sampled values are stored and then formatted into a packet and transmitted to a remote receiver, whereupon they are passed to a remote processing device. The remote processing device may be a Cloud based server. In this example, steps 301-307 are performed by the local processing device onboard the mixing vehicle, and the following steps 308-315 are performed by a remote processing device. In step 308, the packets of data are received by the remote processing device. In step 309, as the packets of data are received, they are cached until a ‘customer docket’ has been received by the system. A customer docket is a set of information, which may be stored or transmitted to the remote processing device. It may include various information necessary for calculating slump, such as the batch information for the concrete loaded into the concrete mixing vehicle. Such a docket may not be immediately available and may only become available a short time after the vehicle has set off to a delivery site. In step 310, a slump calculation model is selected from a set of stored slump calculation models, based on the slump type and other information provided by the customer docket. In step 311, once the customer docket containing the batch information is received, the relevant cached data is retrieved and, together with ongoing received packetised data from the truck, is used to calculate the slump from the time of batching (i.e. when the concrete was first loaded into the drum) and then in real time going forward. In other words, the cached data is processed according to the steps of 312-315 as if it were data arriving in real time. Steps 310 and 311 may be performed only as an initialising step, rather than being performed periodically every time a new packet is received. In step 312, the average and standard deviation for drum torque are calculatedfrom the torque value the average pressure , ) and standard deviation ,pressure outputs provided in the packet received from the vehicle. The calculation of torque from pressure may use a linear relationship, with parameters given by the manufacturer of the hydraulic motor. In some examples, step 312 is performed by the local processing device onboard the concrete mixing vehicle, such that the average and standard deviation for drum torque are generated onboard the concrete mixing vehicle and then transmitted in the packet to the remote receiver. In step 313, the drum speed value (in this case ,) and torque value (in this case ) are processed to calculate one or more vehicle parameters using a relationship estimator Bayesian filter. In this example, two vehicle parameters are estimated, referred to as gradient and offset . The one or more vehicle parameters may be dimensionless values, which do not directly correspond to any physical characteristic of the vehicle or concrete. An exemplary relationship and set of vehicle parameters modelled by a relationship estimator Bayesian filter will be described in relation to Figures 4 and 5. In this example, the relationship estimator Bayesian filter is a Kalman filter, and has a process noise set to a predetermined constant, non-zero value. This value may be determined experimentally and may be considered to represent how much the actual slump of the concrete is expected to change randomly from moment to moment. For each of the two input values and , the relationship estimator Bayesian filter may be provided with an a measurement noise value. In this example, for torque , the measurement noise is proportional to the torque standard deviation . A higher torque standard deviation is suggestive of a higher measurement noise, and vice versa. For drum speed the measurement noise value is a predetermined constant, non-zero value. In step 314, an estimated slump value is calculated by providing the vehicle parameters as an input to a slump calculation model. In this example, each slump prediction model comprises a function which has parameters that have previously been determined by a regression using experimentally generated data. As previously discussed, the slump prediction model is selected 310 corresponding to the type of concrete loaded, as informed by the customer docket. The governing equation of each slump prediction model may be a linear function of the form: =( × offset ) + ( × gradient ) + (1)where A, B, C are experimentally derived, non-zero constants. Alternatively, the governing equation may be any other suitable function, e.g. an equation of the form: =(offset , gradient )In some examples, the slump prediction model also uses additional information to estimate slump. For example, the volume of material in the mixer drum may be used. Alternatively or additionally, the temperature of the concrete in the mixer drum may be measured and this may also form an input to the slump prediction model. In step 315, the estimated slump value is stored and supplied to the customer through an API or user interface. Figure 4 illustrates processing elements of a system 400 configured to generate a slump estimate. The system 400 comprises initial processing elements 401-404, 411- 412 which are implemented by one or more processing units on-board a concrete mixing vehicle (such as e.g. the local processing device 110), and secondary processing elements 421-423 which are implemented by one or more remote processing units (such as e.g. the remote processing device 220). The system may be configured to perform the method described in relation to Figure 3. On-board the concrete mixing vehicle, there comprises a rotation sensor 401. In this case, the rotation sensor 401 is an accelerometer as previously described in relation to Figure 3. The rotation sensor 401 senses acceleration along the two axes Ax, Ay perpendicular to the axis of rotation of the drum. These accelerations Ax, Ay are passed to an angle converter 402, which combines them into a single measurement of drum angle by identifying the direction of the gravity vector in the accelerometer readings. The angle converter passes a measured drum angle M to a drum estimator filter 403, which is a Bayesian filter configured such that the drum rotational speed comprises a hidden state variable. The estimated drum rotational speed is outputted and passed to a smoothing filter 404. The smoothing filter may a low-pass filter, such as a sliding-window averaging filter. Other low-pass filters are known in the art. In an example, the cutoff frequency of the low-pass filter is between 0.5. and 5Hz. Onboard the concrete mixing vehicle there also comprises one or more pressure sensors 411, which are configured to sense the pressure in the hydraulic circuit driving the hydraulic motor which turns the mixing drum. In this example, the pressure sensors 411 comprise one sensor mounted upstream of the drum and one mounted downstream of the drum. The readings P1, P2 of these pressure sensors 411 are passed to a smoothing filter 412. In this example, the smoothing filter comprises a sliding window filter. Thesmoothing filter 412 is configured to calculate the mean , and standard deviation, for each pressure sensor. These values, along with the smoothed estimated drum speed , are then transmitted to the remote location for further processing. A torque 421 calculates a mean and standard deviation for the torque using the pressure values. The mean torque and smoothed estimated drum speed are provided as inputs to a relationship 422. The relationship estimator estimates the values of the gradient and offset vehicle parameters, and comprises a Bayesian filter. The vehicle parameters and are passed to a slump prediction model 423, which uses them to calculate a slump prediction S. The slump prediction model 423 may use a linear equation to calculate the slump prediction S, as previously described. An exemplary relationship modelled by a relationship estimator Bayesian filter is described in relation to Figures 5 and 6. It is known in the art that a slump estimate can be calculated from an instantaneous reading of speed and torque / hydraulic pressure. However, using a single value in isolation is unreliable, as the estimated slump value may be thrown off by an anomalous reading. Figure 5(a) shows corresponding values 501 of drum speed and torque plotted against each other on a graph of drum speed . The values 501 are measured either manually or using sensors on a concrete mixing vehicle, from a mixing drum containing concrete which has a constant slump value S. A line of best fit 502 can be plotted over the measurements 501. The line of best fit may be characterised by a gradient 502A, and an offset 502B where it meets the Y-Axis. In other words, the line of best fit is a linear relationship between torque T and drum speed , modelled by the equation: =+ (2)where is the gradient 502A and is the offset 502B. Figure 5(b) illustrates how the values of the gradient and offset are indicative of the slump value S. Line 503 represents the line of best fit for drum speed and torque values 501 corresponding to concrete in a mixing drum with a relatively low slump value S1. Line 504 represents the line of best fit for drum speed and torque values 501 corresponding to concrete in a mixing drum with a relatively high slump value S2. As the slump value increases and the concrete becomes less “wet”, the amount of torque T required to turn the mixing drum at a given speed increases. As a result, the offset will increase and / or the gradient will become steeper. This can be seen in Figure 4(b), where the offset 504B of line 504 is greater than the offset 503B of line 503, and the gradient 504A of line 504 is steeper than the gradient 503A of line 503. Hence, by estimating the values of and , these values can be used to calculate a slump estimate. Since and model a relationship across a range of drum speed and torque values, they can remain constant even as speed and torque fluctuate. As a result, it is not necessary for the speed and torque to remain constant in order to estimate the values of and . In fact, a degree of fluctuation in torque and speed is not just acceptable but advantageous to spread out measurements across the graph and enable a line of best fit to be plotted. Furthermore, the parameters and of the linear relationship between drum speed and torque hold more information about the rheology of the concrete than a slump test or the pressure alone (e.g. the plastic viscosity). This results in a more accurate slump estimate. As shown in Figure 6, an exemplary relationship estimator Bayesian Filter may be used to model the linear relationship of equation (2), and thereby estimate the values of and . In this example, the relationship estimator Bayesian filter is a Kalman filter, which comprises equation (2) as its internal model, and , as the vehicle parameters and hidden state estimates. At each timestep, the relationship estimator Bayesian filter begins with an initial estimate of and . These initial estimates were the outputs of the filter at the preceding timestep. These values can be represented on a graph of torque T against drum speed , as modelling a line with gradient 601A = and offset 601B = . The relationship estimator Bayesian filter then makes an a priori prediction, in which it extrapolates the vehicle parameters forward in time from the preceding to the current timestep. In this example, the relationship estimator Bayesian filter is configured to predict a priori that slump will decrease over time as the concrete hardens, and thus the values of the gradient 602A and offset 602B increase compared to the preceding timestep’s estimates 601A, 602B. The average torque value and average drum speed value for the current timestep are provided to the relationship estimator Bayesian These values can be represented graphically by an input representation 604 on the graph. The centre of input representation 604 is positioned at the location of the torque and drum speed values on the graph. The uncertainty associated with each of these values is represented by the height and width of the input representation 604, with the height being equal to the torque standard deviation and the width being equal to the drum speed standard deviation . As a result, the input representation 604 is centred on the current (noisy) measurement of torque and drum speed but is also likely to encompass the actual true values. The relationship estimator Bayesian filter combines the inputs (average torque value and average drum speed value ) with the a priori prediction, to generate an updated estimate of the vehicle These are represented by the line 603, which has a gradient 603A = and offset 603B = . These updated vehicle parameter estimates are provided to the slump prediction model, and stored for use in the next timestep. It will be understood that Figure 6 is a graphical representation of the relationship estimator Bayesian filter. The filter itself can perform the steps described above purely mathematically. An exemplary method 700 for generating a slump estimate during discharge modes is now described in relation to Figure 7. The direction of drum rotation is reversed during discharges. Two types of discharge can occur. The first type is a “sample pour”, in which an insignificant volume of concrete is removed from the drum e.g. for manual testing. A sample pour lasts for only a short amount of time (e.g. 30 seconds) and involves a small number of reverse drum rotations. The second type of drum rotation is a “full discharge”, in which a large portion of the concrete in the mixing truck is discharged. A full discharge involves a prolonged period of reversal (e.g. several minutes), and a large number (e.g. 20) of reverse drum rotations. The method 700 ensures that the drum speed and torque values are not utilized during discharges, during which the drum speed and torque sensors may be unreliable. Instead, the slump estimate is generated as a prediction from the relationship estimator Bayesian filter’s internal model. By generating this prediction, calculation of slump from the drum speed and torque values can be resumed following a sample pour. In step 701, drum speed and torque values are received, as previously described in relation to step 31 and 308 of Figure 3. In step 702, discharge is detected. This is done by monitoring the received drum speed value. During charging the drum speed has a positive value, but when the drum rotation direction reverses for discharging, the drum speed becomes a negative value. If the drum speed value is positive, then the drum is charging. Slump is calculated 702 as previously described, for example following steps 32-34 or 310-315 of Figure 3. If the drum speed value is negative, then it is deduced that discharge is occurring. In this case, in step 704 the relationship estimator Bayesian filter is switched from its normal operation to a so-called prediction-only mode. In the prediction-only mode, the relationship estimator ignores the inputs (i.e. average torque value and average drum speed value ), and returns as its output the a priori prediction of the vehicle parameters. In step 705, the predicted values of the vehicle parameters from step 704 are passed to the slump prediction model, and used to generate a slump prediction. Slump estimates produced in prediction-only mode are stored, but in some cases may be marked as invalid so that they are not supplied to the customer. Instead, the slump value supplied to the customer during and following a full discharge may be the slump estimate from immediately before the full discharge began (i.e. corresponding to the last calculation interval where the received mean drum speed was positive). While the drum speed is negative, the system begins counting the number of complete negative rotations, in order to determine whether the discharge is a sample pour or a full discharge. Once a certain number of negative rotations have been counted, the system assumes that a full discharge has occurred and ceases calculating slump estimates. In step 706, it is determined whether a full drum rotation has occurred, e.g. by determining whether the drum angle has increased past a multiple of 360 degrees. If this is the case, then in step 707 the negative drum rotation counter is incremented by 1. In step 708, it is determined whether the prerequisite number of negative drum rotations have occurred. If this is the case, then in step 709 the system records that a full discharge is in progress, and ceases generating slump estimates. In the event of a sample pour, the relationship estimator Bayesian filter switches 704 to prediction-only mode and the predictions of the vehicle parameters are used by the slump prediction model to predict a slump estimate. The system continues to predict 705 the slump estimate until the sample pour is over and the drum begins charging, whereupon the drum speed returns to a positive value and the relationship estimator Bayesian filter returns to its normal mode (i.e. calculating slump per step 703).

Claims

CLAIMS 1. A computer-implemented method for calculating an estimate of a parameter characterising concrete in a drum of a concrete mixing vehicle, comprising the steps of: receiving a drum speed value indicative of a rotational speed of the drum; receiving a torque value indicative of a torque required to turn the drum; providing the drum speed value and torque value as input measurements to a relationship estimator Bayesian filter; calculating at least one vehicle parameter characterising a relationship between drum speed and torque using the relationship estimator Bayesian filter, wherein the relationship estimator Bayesian filter is configured such that the at least one vehicle parameter comprises a hidden state variable of the relationship estimator Bayesian filter; and, calculating an estimated parameter value by providing the at least one vehicle parameter as an input to a parameter calculation model.

2. The method of claim 1, wherein the relationship estimator Bayesian filter comprises a Kalman filter.

3. The method of any preceding claim, wherein the relationship estimator Bayesian filter comprises a non-linear model for modelling the relationship between drum speed and torque.

4. The method of any preceding claim, further comprising the steps of: receiving batch information for the concrete in the drum of the concrete mixing vehicle, and, selecting, based on the batch information, the parameter calculation model from a set of stored parameter calculation models.

5. The method of any preceding claim, wherein the inputs to the parameter calculation model additionally comprise one or more of: the volume of concrete in the drum; and the temperature of the concrete in the drum.

6. The method of any preceding claim, wherein the drum speed value represents an average drum speed over a time period and the torque value represents an average drum torque over said time period.

7. The method of claim 6, further comprising the steps of: generating a standard deviation value representing one or more of the drum speed standard deviation over said time period, or the torque standard deviation over said time period, calculating, from the standard deviation value, a measurement noise value, and providing the measurement noise value as an input to the relationship estimator Bayesian filter.

8. The method of any preceding claim, wherein: the drum speed value and torque value are received at a first periodic interval; the received drum speed value and torque value are sampled at a second periodic interval which is slower than the first periodic interval, to generate a sampled drum speed value and torque value; and the one or more vehicle parameter and the calculation of the estimated parameter value are calculated at the second periodic interval using the sampled drum speed value and torque value.

9. The method of any preceding claim, further comprising the steps of: receiving a rotation measurement from a rotation sensor and a torque measurement from a torque sensor; and, processing the received rotation measurement and torque measurement to respectively generate the drum speed value and torque value.

10. The method of claim 9, wherein processing the received rotation measurement comprises the steps of: providing the rotation measurement to a drum estimator Bayesian filter; calculating the drum speed value using the drum estimator Bayesian filter, wherein the drum estimator Bayesian filter is configured such that the drum speed value comprises a hidden state variable of the drum estimator Bayesian filter.

11. The method of claim 10, wherein the drum estimator Bayesian filter comprises a Kalman filter.

12. The method of claim 10 or 11 when dependent on claim 7, further comprising the step of calculating the drum speed standard deviation using the drum estimator Bayesian filter, wherein the drum estimator Bayesian filter is configured such that the drum speed standard deviation comprises a hidden state variable of the drum estimator Bayesian filter.

13. The method of any of claims 9-12, wherein processing the received torque measurement comprises the steps of: providing the received torque measurement as an input to a sliding window averaging filter, and recording as the torque value an output of said filter.

14. The method of claim 13 when dependent on claim 7, further comprising the step of recording the torque value standard deviation as an output of the sliding window averaging filter.

15. The method of any of claims 9-14, further comprising the steps of: transmitting, from a local transmitter, the drum speed value and torque values; and receiving, at a remote receiver, the drum speed and torque values.

16. The method of any preceding claim, further comprising the steps of: determining, from the drum speed value or drum torque value, whether the concrete mixing vehicle is in a discharge mode, and if the vehicle is in a discharge mode, switching the relationship estimator Bayesian filter from a normal mode to an alternate mode.

17. The method of claim 17, wherein the relationship estimator Bayesian filter comprises a weighting value for weighting the influence of the drum speed value or torque value as input measurements, and the weighting value is lower in the alternate mode compared to the normal mode.

18. The method of claim 18, wherein the weighting value is zero in the alternate mode.

19. A system for estimating a parameter characterising concrete in a drum of a concrete mixing vehicle, comprising a processor configured to carry out the method of any of claims 1-8.

20. The system of claim 19, wherein the processor is a remote processor.

21. The system of claim 20, further comprising: a local processor;a local transmitter; a remote receiver; and, wherein the processor is configured to: receive a rotation measurement from a rotation sensor and a torque measurement from a torque sensor; and, process the received rotation measurement and torque measurement to respectively generate the drum speed value and torque value, and wherein the local transmitter is configured to transmit the generated drum speed value and torque value to the remote receiver, and wherein the remote receiver is configured to receive the drum speed value and torque value and communicate them to the remote processor.

22. A non-transitory computer-readable storage medium storing instructions to cause the processor of claim 16 to perform the method of any of claims 1-8.

Images