Search

Signal processing is a fundamental component of almost any sensor-enabled system, with a wide range of applications across different scientific disciplines. Time series data, images, and video sequences comprise representative forms of signals that can be enhanced and analysed for information extraction and quantification. The recent advances in artificial intelligence and machine learning are shifting the research attention towards intelligent, data-driven, signal processing. This roadmap presents a critical overview of the state-of-the-art methods and applications aiming to highlight future challenges and research opportunities towards next generation measurement systems. It covers a broad spectrum of topics ranging from basic to industrial research, organized in concise thematic sections that reflect the trends and the impacts of current and future developments per research field. Furthermore, it offers guidance to researchers and funding agencies in identifying new prospects., 48 pages, https://iopscience.iop.org/article/10.1088/1361-6501/ac2dbd

The Industrial Internet of Things (IIoT) and Industry 4.0 require new intelligent sensor designs with enhanced functionality, including local diagnostics. In previous work, we have experimentally investigated an important fault mode of a commercial pressure sensor, working in partnership with the sensor manufacturer who has provided modified sensors with calibrated levels of the fault condition. We have further developed simple signal processing techniques to detect the fault condition, based on a low cost noise analysis. In the current paper, we describe the development of a prototype wireless pressure transmitter. This transmitter monitors the analogue output of the pressure sensor, and applies the diagnostic procedures in real time. The resulting pressure measurement in engineering units, together with diagnostic information, are both communicated wirelessly to a receiving system.

The Prism is a new signal processing module which implements fully recursive, linear phase FIR filtering, so its computational cost is fixed irrespective of filter length. The Prism also has negligible design cost. Recent work has demonstrated how, using simple design rules, a chain of six Prisms can create a narrowband filter with arbitrary central frequency and bandwidth. In this paper, the technique is applied to spectral analysis of data, whereby a sequence of filters is applied to a data set to provide narrow frequency analysis, yielding accurate estimates of the frequencies and amplitudes of spectral peaks in the data. Although this time domain technique remains computationally expensive compared to the FFT, it can identify and reject spectral leakage, offering an alternative analysis for low amplitude and/or adjacent spectral peaks, including hidden tones, where FFT discrimination may be limited.

Prism signal processing is a new FIR signal processing technique offering a fully recursive calculation and minimal design cost, which is well-suited to the requirements of IoT sensing. This paper describes a family of low pass filter designs using one or more layers of Prisms, where the frequency response has minimal stopband gain. Combined with heterodyning and Prism-based tracking, these filters provide flexible and powerful real-time tracking of signal component properties for use in measurement applications.

This paper presents new complex signal processing methods, Complex Bandpass Filtering (CBF), Complex Notch Filtering (CNF) and their combination (CBF-CNF), applied to the Coriolis Mass Flowmeter (CMFM). CBF, CNF and CBF-CNF can be utilized to suppress the negative frequency component of each sensor signal in order to produce the corresponding analytic form with reduced tracking delay. Further processing of the analytic form yields the amplitude, frequency, phase and phase difference of the sensor signals. In comparison with previously published methods, the new techniques offer short delay, high noise suppression, high accuracy and low computational cost. A reduced delay is useful in CMFM signal processing especially for maintaining flowtube oscillation in two/multi-phase flow conditions. The central frequency and the frequency range for the CBF, CNF and CBF-CNF methods are selectable so that they can be customized for different flowtube designs.

This paper introduces the Prism, a new type of signal processing block, as a contribution to the challenges of 21st Century metrology. The Prism is a fully recursive, dual output, FIR filter: the computational burden is low and independent of data window length. Prism design is trivial, so that networks of Prisms can be created, whether at design time or autonomously in real time, to carry out a range of metrological tasks. Prism-based trackers can generate sample-by-sample estimates of frequency, phase, and/or amplitude of a sinusoid. A simulation example of sensor validation demonstrates how Prism signal processing can be used to autonomously detect, track, and compensate for an undesired frequency component in a frequency-based sensor.

A new, networked approach to signal processing, efficient for both design and real-time computation, is proposed to match the requirements of the Internet of Things (IoT). The Prism is a novel finite impulse response (FIR) filter, which is fully recursive: the computational cost per sample is low and independent of the filter window length. The design cost is also low; any device capable of running a Prism-based signal processing scheme may also adapt the design to match changing requirements. Various signal processing tasks can be implemented using networks of Prisms, thereby providing a useful tool set for a wide range of sensing and monitoring applications, including the IoT.

— Two-phase (gas/liquid) flow is common in many industrial applications but its measurement remains challenging for Coriolis mass flow meters (CMFM), especially for high Gas Void Fraction (GVF). In this paper, we present experimental results applying previously developed complex signal processing techniques for tracking the rapidly changing sensor signals generated by two-phase flow. The techniques are implemented in a new System-on-Chip (SOC) prototype transmitter connected to a commercial Coriolis flow tube. Experiments have been carried out over a range of single phase and two-phase (water/air) flows. The signal tracking as well as mass flow rate and density measurement performance is compared with that of a laboratory version of a commercial Coriolis transmitter with two-phase flow capability. The results show that the complex bandpass algorithms, coupled with flowtube control algorithms, reduce the standard deviation of the mass flow measurement by a factor of 3 or more in 50% of the experiments undertaken. For the density measurement, the corresponding reduction in standard deviation is by a factor of 6.

Knowledge of fuel mass injected in an individual cycle is important for engine performance and modeling. At the moment, such measurements are not possible on engine or in real time. In this article, a new method using Coriolis flow meters (CFMs) and a new, patented, signal processing technique, known as the Prism, are introduced. CFMs are extensively used for flow measurement both in the automotive industry and further afield and, when coupled with the Prism, have the potential to make these challenging high-speed measurements. A rig-based feasibility study was conducted injecting very small quantities of diesel (3 mg) at pressures of up to 1000 bar at simulated engine speeds of up to 4000 rpm. The results show that these small quantities can in principle be measured. The results also reveal a previously unknown behavior of CFMs when measuring very low flow rates at high speed. The study concludes that by combining high-resonant frequency flow tubes with the Prism technology in a new instrument—the fast next-generation Coriolis (fast NGC) flow meter—it will be possible to measure individual injector flow rates on engine in real time.

The verification of the proper operation of individual sensors is increasingly important as a means of validating the measurements that they generate. One technique for verifying the structural integrity of a transducer and its housing is to use ultrasonic probing, whereby an ultrasonic pulse is generated within the sensor housing. The response signal, which contains multiple resonances, indicative of structural health, is monitored and analyzed. Alternative techniques have been applied to analyze the ultrasonic response signals, obtained from a pressure sensor housing. These include Prism Signal Processing and the Matrix Pencil Method. Both of these techniques are able to track the frequency and amplitude of six or more modes of vibration. Experimental results demonstrate that large shifts in calculated parameter values occur when a fault condition is introduced. In addition, the time varying amplitudes of the modes do not always exhibit the expected exponential decay: this characteristic behaviour may provide an important additional diagnostic information.

The Xilinx Artix-7 family of FPGAs has a peak DSP performance of approximately 1 TMAC/s. This paper describes a floating point FIR bandpass filter instantiation on an Artix-7 with up to 192 million taps, updated at up to 2 MHz, delivering an equivalent performance of up to 384 TMAC/s, while using only one third of the DSP resources on the FPGA. This has been achieved using Prism signal processing, a new recursive FIR filtering technique. Filter design is simple; a PC host HMI facilitates the immediate creation of new filter implementations. Examples of experimental results are provided.

An earlier study described a wireless acceleration sensor used to perform condition monitoring of rotating machinery, via off-line spectral analysis. In this paper, a new technique called Prism signal processing is applied to the same problem to provide sample-by-sample analysis of the machine rotation. The Prism has low design cost and low computational cost, and so is particularly suited to the flexible signal processing requirements for the Internet of Things. This paper provides an overview of the sensor and its experimental rig as well as the Prism-based signal processing scheme used to track the key components for diagnostic monitoring. The validity of the signal processing scheme is demonstrate via a simulation.

The Internet of Things (IoT) concept, alongside wireless technologies, supports the mounting of sensors in inaccessible positions and thus provides new opportunities for machine condition monitoring. This paper outlines experimental results using Prism signal processing to track rotor angular acceleration via a Wireless Acceleration Sensor (WAS) mounted on a rotating shaft. The instantaneous frequency and amplitude of each component of the angular acceleration is tracked, with a view to providing diagnostic information. The experimental results illustrate how amplitude data can provide indications of gear faults, via further Prism signal processing.

Our companion conference paper describes the requirement for flexible, self-validating instrumentation to support the goals of Industrial Cyber-Physical Systems. For high-value instruments such as Coriolis Mass Flow (CMF) meters, this entails more sophisticated signal processing techniques to provide additional sources of diagnostic and measurement data, particularly in dealing with common, but difficult real-world conditions such as two-phase (gas/liquid) flow. A novel Matrix Pencil Method (MPM) was described, adapted to monitor two modes of vibration for the CMF flowtube, and its performance evaluated in simulation. In this paper, we present results from applying the MPM technique to experimental data.

This paper introduces a new technique, Prism signal processing, which may be used for the tracking of one or more noisy sinusoids in a signal. A simulation study is presented demonstrating the potential of Prism signal processing as an alternative to Prony’s method for analyzing exponentially decaying sinusoids. One application is to sensor condition monitoring of an industrial pressure sensor, using ultrasonic excitation to evaluate the sensor’s structural integrity. Initial experimental results suggest the Prism technique can reveal details in the resulting frequency/amplitude time series of each component, which is not available through Prony’s method.

The Prism is a new signal processing object which provides fully recursive FIR filtering i.e. the computational load is small and fixed irrespective of filter length. The Prism also has negligible design cost. These features are attractive to the next generation of Industry 4.0 sensors requiring low cost and flexible signal processing. Recent work has demonstrated how, using a few simple design rules, a chain of six Prisms can create narrowband filters with arbitrary central frequency and bandwidth, also with low design and computational cost. In this paper this technique is extended to provide spectral analysis of data, whereby a sequence of filters is applied to a data set to provide selective frequency analysis.

This paper explores the possible application of the complex bandpass filter in Coriolis mass flow meter signal processing. The method shows high robustness for frequency, amplitude and phase tracking under noisy conditions, such as those generated by two-phase (gas/liquid) flow. Simulations of the new technique are compared with existing academic and industrial signal processing solutions to evaluate their relative performance.

Improved signal processing methods for Coriolis Mass Flow Meters provide more accurate measurements and offer a wider range of applications, such as two-phase flow. This paper reviews current signal processing methods for amplitude, frequency and phase difference tracking and analyses their performance in a simulation of two-phase flow conditions.

The dynamic response of flow meters is significant in many applications, including fast control operations, e.g. short duration (less than 1 s) batch filling and for tracking the periodic flow fluctuations produced by positive displacement devices. The factors which determine Coriolis dynamic response have been elucidated. It has been shown that the meter flow tube response time cannot be less than the duration of one drive cycle of the tube vibration (i.e. reciprocal of drive frequency). This gives the potential of a response time of order 1 ms for the fastest currently available meters. However, the delay-time and update rates from the user output depend upon flow transmitter technology and design. Flow tube dynamic response has been investigated theoretically (simple straight tube), by finite element simulation (complex flow tube shapes) and experimentally. Commercially available meters were tested to determine the flow tube dynamic response to step changes in flow rate and the response to low frequency (compared with meter drive frequency) flow pulsations. Generally, dynamic flow events have been found to introduce contaminating signal components at one or more frequencies, other than that of the meter drive. The paper also presents details of the signal processing used to extract the required phase-difference and a method for reducing the contaminating signal noise. A new fast-response meter is currently being developed and some of the significant advances in the technology of a novel digital transmitter are described.

This paper discusses Small Volume Provers (SVPs), used in the oil and gas industry to validate the performance of custody transfer meters. Recently Coriolis mass flow meters have been introduced for custody transfer; while these offer reduced maintenance requirements over traditional PD and turbine meters, proving Coriolis flow meters using SVP is challenging. This paper presents SVP results for a Coriolis meter which matches or exceeds the most stringent requirements for custody transfer. This is achieved in part by using a novel signal processing technique which reduces the dominant component of the measurement noise, associated with so-called Coriolis mode vibration, with negligible loss of dynamic response. © 2006 Elsevier Ltd. All rights reserved.

A previous paper (Henry, 1995a) introduced the technique of hardware compilation as the basis for developing highly flexible programmable hardware platforms for control applications such as sensor validation. This paper describes two PC-hosted architectures for sensor validation research. The first holds up to two FPGAs and supports a daughter board with application-specific circuitry. The second is based on the transputer TRAM standard, and consists of programmable hardware modules providing interfacing and low-level signal processing between the transputer and arbitrary I/O components. Three applications are described, based upon a thermocouple, a dissolved oxygen probe and a Coriolis mass flow meter.

Recent years have seen significant advances in Coriolis flowtube technology Arguably, however, it is advances in transmitter technology that will ultimately prove more significant in providing users with the desired level of measurement performance and robustness. The key to transmitter enhancement is new digital technology, enabling improvements in flowtube control, signal processing, and measurement quality. The Sensor Validation Research Group at the University of Oxford, in partnership with Invensys Foxboro, has been developing Coriolis transmitter technology for many years. The latest Oxford prototype is the basis for the new Foxboro CFT50 Coriolis transmitter product. Key features include digital synthesis of the drive waveforms, and maintaining operation during two-phase or aerated flow.

The article provides an overview of the techniques that can be used for condition monitoring of railway equipment, using as an illustrative example a points validation system. Aspects to be covered include fault analysis, fault coverage, sensor selection, signal processing, fault modelling, and communications. (20 pages)