Flexible flow sensors-enabled intelligent life

In our daily life, flexible flow sensors endow us with a ‘sixth sense’ capability, i.e. ‘touch’ the fluids, improving living quality. Although there are kinds of flexible flow sensors developed to implement this capability, they still have insufficient sensitivity and limited intelligent applications in daily life. Biomimetic engineering provides us with a powerful and effective approach to develop highly sensitive and intelligent flow sensing systems served in our life, comparable to that in creatures. Here, in this review, we present a comprehensive review of recent studies on the flexible flow sensors for human intelligent life. Firstly, we briefly introduce the excellent flow sensing systems selected by nature, and typical design strategies of artificial flexible flow sensors. Furthermore, we collect and exhibit kinds of flexible flow sensors and their applications in intelligent and digital life. Finally, we discuss the challenges and future perspectives of the flexible flow sensor for the metaverse applications.


Introduction
Metaverse exhibits a self-sustaining, hyper spatiotemporal, and 3D immersive virtual shared space, realized by the Original content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence.Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.convergence of physically persistent virtual space and virtually enhanced physical reality [1].Internet of Things (IoT) is a network composed of hundreds of physical objects, which are embedded with sensors, communication components, softwares, and other technologies.Through inter-connected smart devices for digitalization, IoT bridges the physical world and the digital world, ensuring the information flowed freely between these two worlds [2].As a cutting-edge technology of the metaverse, IoT has been well developed by rapidly developed sensors and internet technologies [3,4].Furthermore, it realizes an intelligent life style involved in various aspects of our daily life [5,6], including navigation, healthcare, city management, agriculture, etc.
As an interface between human and machine, flexible sensors play crucial roles in the metavase [7], due to their numerous advantages, such as flexibility, lightweight, etc.Recently, there are various flexible sensors developed for the application in the metavase, including flexible tactile sensors [8], soft olfactory sensors for virtual response [9], conformal strain sensors for sign language expression [10], which achieve splendid human-machine interaction.For instance, Yu et al developed a wireless, battery-free electronic systems and touch-based haptic interfaces which could be integrated onto skin curved surface to communicate information via spatiotemporally programmable patterns of localized mechanical vibrations [8].The resulting haptic interfaces provided many opportunites for usage, where the skin offered an electronically programmable communication and sensory input channel to the body, with typical demonstrations including applications in social media and personal engagement, prosthetic control and feedback, and gaming and entertainment.However, besides these common stimuli, flow field perturbation is a fundermental stimulus in our daily life.It can be said that we live in fluids, including air and water.Biomimetic engineering provides us with an effective approach for the development of highly sensitive flexible electronics [11][12][13][14][15]. Inspired by the natural flow sensing systems, researchers have proposed kinds of flexible flow sensors to render human with a 'sixth sense' capability to interact with the real world [16,17].The 'sixth sense' herein means that human can recognize surrounding environments by decoding flow field information, without using the traditional five senses of sight, hearing, touch, smell, and taste.
In this review, we summarize the potential applications of flexible flow sensor-assisted IoT for the 'sixth sense' implementation, as illustrated in figure 1.Firstly, flow sensing strategies in nature and flexible artificial counterparts in engineering are reviewed.Consequently, the potential applications for various intelligent scenarios are demonstrated based on flow sensing.In final, the underlying challenges and outlook of flexible flow sensors for IoT applications are discussed.

Biological strategies
In nature, animals have evolved excellent sensing systems for hydrodynamic or aerohydrodynamic sensing, through natural selection [16,17].For instance, fish and seals have unique lateral line systems (LLS) and specific whiskers for hydrodynamics perception, assisting their survival in dark or cloudy environment, respectively.Spiders and crickets can escape and prey, benefited from their highly sensitive cilia organs.These remarkable mechanosensory systems provide engineers with prototypes for developing counterparts serving the human community.Here, we present a brief introduction of the sensing performance and sensory structures of these systems at first.
With the assistance of mechanosensory LLS, fish can carry out a series of flow field-based behaviors, such as, rheotaxis, prey and predators, even in blind scenarios [16].The LLS is composed of hundreds of functional units, termed as neuromasts, distributed across the fish body (figure 2(a)(i)) [26,27].The sensing mechanism of the neuromast is described as: the water movements bend the cupula, then activate or inhibit the ion channels in hair cells (figure 2(a)(ii)).Canal neuromasts (CNs) and superficial neuromasts (SNs) are two categories of LLS, based on their spatial positions (figure 2(a)(iii)) [28].Freestanding on the skin surface, the SNs are sensitive to flow velocity, and can detect a magnitude of µm s −1 .CNs are located in subepidermal fluidfilled microchannels connected to the external environments through series of pores, which perceive pressure difference between adjacent canal pores reaching magnitudes on the order of mPa [29].
Harbor seals are able to forage in low visibility environments, and even at depths greater than 300 m.Their whiskers allow them to follow the vortices left by prey and detect objects that passed by even 30 s earlier (figure 2(b)(i) and (ii)) [30].Remarkably, the whiskers of harbor seals have a peculiar cross-sectional shape: the cross-section is elliptical, and the ratio of the major and minor axes varies along the whisker, with a period of approximately 1-3 mm, thus giving the whisker an undulated surface structure that distinguishes it from the whisker of eared seals (figure 2(b)(iii)) [31].Through computational fluid dynamics (CFD) simulation, a vortex core reconstruction from one instantaneous time-step for a vibrissa (top) was revealed, in comparison with those from a circular and an elliptic cylinder (figure 2(b)(iv)) [19].The flow fields behind the vibrissa were greatly different from that behind the comparison cylinders: there was no large-scale primary vortex structures in the wakes of the vibrissa; the vortex structures in the wake flow field for the vibrissa were unstable over a significant downstream distance.The underlying mechanism was uncovered as: Kármán vortices were replaced by a complex vortex structure in which various vortices are generated simultaneously in the direction of the whisker axis.The results suggest that seal whiskers are highly sensitive as flow field sensors, especially for changing eddy current signals.
Filiform hairs presented on the cricket's cerci (figure 2(c)(i)) are known as the most sensitive biological flow sensors [32].To elicit an action potential, the work needed to drive the hair shaft to deflect enough is on the order of 10 −21 J for a cricket filiform hair, which corresponds to 1/100 of the energy of a single photon [33].Spiders not only have lyriform slit organs [11], which enables ultrasensitive displacement detection by allowing for mechanical compliance to detect small external force variations in their surroundings, but also have trichobothria for flow sensing (figure 2(c)(ii)) [34].Both the filiform hairs presented on the cricket's cerci and trichobothria appeared on the spider's leg are located in a cup-mouth shape cuticular socket and the hair shaft forms a simple lever with the long arm exposed to the motion of airflow and the short arm embedded in the exoskeleton coupling with the sensory cells (figure 2 An overview of flexible flow sensors for IoT applications enabling intelligent life, including principles of nature inspiration (lateral line system (Reproduced from [16].© IOP Publishing Ltd All rights reserved.),seal whisker (Reproduced from [18].© IOP Publishing Ltd All rights reserved.Used with permission of The Company of Biologists Ltd, from [19]; permission conveyed through Copyright Clearance Centre, inc.), spider filiform and cricket flow sensilla (Reproduced from [20].© IOP Publishing Ltd All rights reserved.)),engineering strategies, sensing mechanisms, and representative applications (intelligent navigation (Reproduced from [21] with permission from Springer Nature.© [2022] IEEE.Reprinted, with permission, from [22].), intelligent healthcare (Reproduced from [23], with permission from Springer Nature.), intelligent city (Reproduced from [24].CC BY 4.0.),intelligent agriculture (Reproduced from [25].CC BY 4.0.).

(c)(iii)). The length ratio
of the outer to inner arm is more than 1000/1, which indicates that the displacement of the hair's tip is scaled down while the force is scaled up. Figure 2(c)(iv) shows the schematic diagram of the flow-sensing hairs [20,35].The inner end of the hair shaft couples with the dendrites and the viscoelastic suspension behaves as a combination of a spring and a dashpot, which resists the deflection of the hair shaft.As for the trichobothria in spider Cupiennius Salei, the spring stiffness is on the order of 10 −11 -10 −12 Nm rad −1 and the damping coefficient is on the order of 10 −14 -10 −15 Nms rad −1 [36,37].These extremely low values result in the pure tilt without bending when suffered from the viscous force of airflow.

Engineering design strategies
Mimicking the natural excellent flow sensing organs, there are three main strategies employed in engineering implementation, i.e. mechanical deformations, thermal-based, ultrasound flow sensing mechanisms.Mechanical strategical design is widely utilized, due to its feasible and convenient integration with different sensing mechanisms.There are four major types of configurations, including solid film, diaphragm, cantilever, and cilium, as illustrated in figure 3(a).The first two types of structures are within boundary layer due to their small thickness, which do not destroy the interaction between (i) Neuromasts distribution on fish body, (ii) illustration of working mechanism in neuromast, iii) two main types of neuromasts.Reproduced from [16].© IOP Publishing Ltd All rights reserved.(b) Harbor seal whiskers for hydrodynamic sensing.(i) Harbor seals, (ii) cross-section of the harbor seals whiskers, (iii) microstructure of seal whisker, (iv) CFD simulation results of the wakes generated by different whiskers.Reproduced from [18].© IOP Publishing Ltd All rights reserved.Reproduced with permission from [19].[insert copyright line, if specified].(c) Sensing systems of spiders and crickets for aerohydrodynamic sensing.(i) Trichobothria appeared on the spider's leg, (ii) filiform hairs presented on the cricket's cerci, (iii) cuticular socket, (iv) schematic diagram of the flow-sensing hairs.Reproduced from [20].© IOP Publishing Ltd All rights reserved.
surrounding flow fields and carrier.It means the sensitivity is low, due to weak fluid-structure interaction (FSI) behavior (i.e. the interaction of flexible sensor structure with surrounding fluid flow).To increase the sensitivity, researchers have developed flexible flow sensors based on microcantilevers.Compared to the film type with fixed boundary, the cantilever type with free end greatly enhanced flow sensing capability.In nature, the biological flow mechanosensory receptors are mainly cilia configuration injecting into flow fields, directly interacting with flow fields outside the boundary layer.This cilia type flow sensors have highest sensitivity, compared with above three types flow sensors working inside the boundary layer.In addition, the morphology of cilia has a great effect on the sensing performance of flow sensors.The rule is that: the larger surface area endows cilia with larger space for interaction with surrounding fluids, then generating larger viscous force, thereby inducing larger mechanical deformation, finally outputting more significant signals in flow sensors [16,17].
The main working principle is can be described as that a sensing element can detect the mechanical deformation induced by the FSI between flow sensors and surrounding flow fields.The common transduction mechanisms are widely used to detect this tiny strain, such as piezoresistive, capacitive, magnetic, optical grating, and self-powered piezoelectric and triboelectric mechanisms, as shown in figure 3(b).The details are exhibited as below: (1) Piezoresistive sensors can convert applied strain on the device into electrical resistance changes [38], as presented in figure 3(b)(i), which can be regulated by: where ρ is the material's resistivity, L is the length, A is the contact area, and R is the contact resistance [39].They have wide applications due to series of advantages, such as wide detection range, simple fabrication process and detection methods, high stability, etc [40].Structural, material, and mechanical strategies are explored to enhance the performance of piezoresistive flow sensors.For instance, a selfbending technique was utilized to create curved microcantilever structures to further improve the sensing performance of flexible flow sensors [21,41].Several material strategies were proposed to prepare highly sensitive piezoresistive composites, including electric-breakdown techniquebased graphene composites [42,43], porous piezoresistive materials [44], directional liquid spreading strategy-based graphene piezoresistors [45] inside which graphene flakes were adopted as conductive fillers.
(2) Capacitive sensors can convert applied strain on the device into capacitance changes, due to the changes in the distance between two electrodes [40], as illustrated in figure 3(b)(ii).Neglecting the edge effect, the capacitance can be calculated by the following equation: where C is the capacitance of the sensor, ε 0 is the space permittivity, ε r is the relative permittivity of the dielectric material, A is the overlapping area of the two electrodes, and d is the separation between the two electrodes [46].Mechanical and material strategies are employed to improve the performance of capacitive flow sensors, for instance, a free-standing single wing integrated with a interdigitated comb finger capacitor [47], difference design in capacitive changes recording [48], liquid metal plates integrated into a soft silicone cupula [49].
(3) The optical sensors can convert applied deformation or strain into the changes of light wavelength and intensity, as exhibited in figure 3(b)(iii).The variation of Bragg wavelength induced by the mechanical stress can be calculated as [50]: where n eff is the effective refractive index of the fiber core within the Bragg grating, Λ is the period of its changes.Morphological strategies are utilized to enhance the performance of optical flow sensors, such as soft polymer-based optical waveguide [51,52], weight attachment at distal end of optical fiber [24,53].
(4) Piezoelectric sensors can convert applied mechanical force into electric potential produced in certain materials [54], as presented in figure 3(b)(iv).Hooke's law is combined with electrical equations benefits the understanding of mechanism of piezoelectric sensors.The output piezoelectric signals induced by the changes in strain can be calculate by the following equations [55]: where the equation for electrical behavior is expressed as where D is the electric charge density displacement, ε is permittivity, and E is electric field strength.The Hooke's law is expressed as: where S is the strain, s is the compliance, and T is the applied force.Material and integration strategies are employed to design the piezoelectric flow sensors, including nanofibrous piezoelectric materials [56][57][58], piezoelectric fiber with a symmetric electrodes metal core [59], and polyvinylidene Fluoride membrane integrated with preamplification circuits [60].By virtue of the volume-phase transition of thermoresponsive hydrogel, a flexible flow sensor based on a piezoelectric microcantilever integrated with a hydrogel cilium was proposed to achieve tunable sensitivity [61].
(5) The triboelectric nanogenerator (TENG)-based sensor can convert mechanical deformations into electrical charge signals through two triboelectric materials charged opposite charges by mutual friction [40], as shown in figure 3 The governing equation of TENGs can be expressed as: where V is the total voltage, C is the capacitance between the two electrodes, X is the distance between these two triboelectric layers, Q is the transferred charges from one electrode to another electrode, and V OC is the polarized triboelectric charges and their contribution to the voltage [62].To enhance the performance of triboelectric flow sensor, materials and structural strategies are wide employed, such as structured polytetrafluoroethylene film [63,64], soft fluorinated ethylene propylene film [65], and a TENG-based underwater bionic whisker sensor for vortices detection [66].(6) The magnetic sensors can convert external force into the variation in magnetic stray field, as illustrated in figure 3(b)(vi).The fundamental transduction mechanism behind the giant magnetoelastic effect can be illustrated by a wave chain model [40].The relationship between the vertical magnetic field H ver and principal stretch λ could be given by: )) (9) where r is the radius of the magnetic particle; a is the aspect ratio of the wavy chain structure; λ is the stretch in the compression direction; M is the magnetization of the micromagnets; k is a constant characterizing the influence of nonideality, neighboring chain-chain interaction and macroscopic shape effect on the demagnetizing factor under compressive deformation; b and h are horizontal and vertical distances, respectively, between the neighboring micromagnets inside the wavy chain; and (0.3006−f (x)) is the dipole alignment factor, which describes the contribution of all other dipoles to the vertical magnetic field of a single dipole in the wavy chain [40].Material and structural strategies are utilized to optimize the performance of magnetic flow sensors, such as iron nanowires/polydimethylsiloxane (PDMS) compositesbased flow sensors [67], PDMS/NdFeB composites-based flow sensors [68].
Besides mechanical deformations, there are kinds of thermal-based flow sensors developed for wide applications in every field, including navigation, healthcare, etc. Thermal flow sensors utilize the heat transfer intensity in order to determine the flow velocity.In general, thermal flow sensors consist of two basic parts, namely heaters and sensing elements, as shown in figure 3(c).A sensing element detects the variation in heat transfer between the heater and the working flow, therefore, the sensitivity of the device improves when more heat is transferred to the working fluid [69].For typical thermal flow sensor materials, the resistance relationship to temperature is given by [70]: where R(T) is the resistance at temperature T and α is the temperature coefficient of resistivity (TCR).TCR can be determined experimentally by: in which a R is the resistance overheat ratio and determined by measuring the change in resistance of the sensing material at two different temperatures.It appears that one of the most important limitations on accuracy applying to the conventional temperature-based flow sensors is maintaining the temperature on the sensing element accurately.Deep reactive ion etching (DRIE) trenches and full-bridges technologies were used to improve sensitivity of thermal flow sensor.DRIE technology was adopted to fabricate insulation trenches between the heater and thermistors to reduce the unwanted lateral heat loss in the chip, and eight thermistors symmetrically placed in four directions around the heater were configured to two Wheatstone full-bridges (an instrument or a circuit consisted of four resistors or their equivalent in series which is used to determine the value of an unknown resistance when the other three resistances are known [71]) to achieve two times the output with respect to four thermistors [72,73].
In addition, ultrasound mechanism is employed to measure the flow fields, especially for the blood flow velocity evaluation.Its working principle can be simplified described as: a ultrasound wave with a specific frequency of f 0 is transmitted from piezoelectric elements into skin; then moving scatterers (for instance red blood cells) feedback an echo with a frequency of f R , with a certain deviation (i.e.Doppler shift f D ) from the transmitted frequency, which can be calculated as: where c is the speed of sound, V is the flow velocity, and θ, known as the Doppler angle, is the angle between the axis of the ultrasound beam and the direction of flow, looking toward the transducer (figure 3(d)).Material and structural strategies are utilized to enhance the performance of ultrasound flow sensors, for instance, high-performance rigid 1-3 piezoelectric composites hybridized with soft structural components [74], a flexible and skin-conforming ultrasound device for the monitoring of blood flow velocity [75], a fully integrated ultrasonicsystem on patch [76].

Intelligent guidance for vehicles
Learned from nature, various flexible flow sensors have been proposed and validated to improve the vehicles guidance intelligence, including underwater unmanned vehicles (UUVs) and unmanned aerial vehicles, only relied on fluid cues.For instance, as exhibited in figure 4(a), Ma et al developed a flexible flow sensors consisted of PDMS microchannels, and inside poly(vinylidene fluoride-trifluoroethylene)/polyimide cantilever as the sensing units.The device was able to recognize the size of underwater obstacles, with a mean predication error as low as 2.5% [26,77].In addition to obstacle's size, its spatial position is crucial during navigation process.To tackle this problem, Jiang et al proposed a dual-sensor (pressure/velocity) modality fusion strategy to precisely locate near-field obstacles through hydrodynamic stimuli decoded by neural network [22].The experiments revealed that the dual-sensor fusion modality had 30% mean localization error less than that in single pressure or velocity perception modality (figure 4(b)).
UUVs require excellent control methodology to achieve precise navigation in sea environments.The controllers of many UUV employed models to predict the significant forces imparted from the surrounding fluid and to introduce additional control terms or adjust control gains in accordance.Inspired by the fish LLS, Krieg et al developed a control methodology to directly evaluate the fluid forces acting on the UUV using a distributed artificial lateral line sensor modules on the UUV surface (figure 4(c)) [78].Compared to a standard position error feedback controller, the experimental results validated that the proportional derivative controller combined with the lateral line feedforward term significantly decreased position tracking errors by 72%.
Besides underwater flow sensing, airflow sensors have been extensively studied for applications in flow field monitoring and active flow control of unmanned aerial vehicles.As shown in figure 4(d), Shen et al presented a novel flexible airflow sensor based on four curved microcantilevers arranged in a cross-form configuration [21].A self-bending method based on MEMS technology was employed to fabricate the curved microcantilever structures, successfully transferring a 2D plane structure into a 3D structure with a good consistency in the morphology.The developed flexible flow sensors can be conformally attached on the surface of the air vehicles, achieving flow field monitoring and angle of attack (α) detection.Que and Zhu fabricated a micro hot-film flow sensor array mounted on the surface of the wing, utilizing flexible materials but with hot-film principles [79], as shown in figure 4(e).A back-propagation neural network was used to model the coupling relationship between readings of the sensor array and aerodynamic parameters.The calculated flight parameters by the sensor measurements fit the actual flight parameters very well.The measurement errors of air speed, angle of attack, and angle of sideslip were approximately 0.27 m s −1 , 0.87 • , and 0.27 • , respectively.Air Force Research Laboratory has developed an artificial hair sensor (AHS) mimicking those used by natural fliers (such as bats, locusts, and crickets) to detect surface flow features [80].The AHS consists of a glass fiber with radially grown carbon nanotube (CNT) forests on it.The surface flow features such as surface flow velocity was measured as the change in resistance between two electrodes due to shaft deflection and subsequent CNT compression.Local flow measurement from an array of AHSs in a wind tunnel experiment was used with feedforward artificial neural network to predict aerodynamic parameters such as lift coefficient, moment coefficient, free-stream velocity, and angle of attack on an airfoil.As shown in figure 4(f), inspired by the wing membrane of a bat, a highly sensitive and adaptive graphene/single-walled nanotubes-Ecoflex membrane (GSEM)-based airflow sensor mediated by the reversible microspring effect was developed [81].Owing to the excellent spatial resolution of the airflow sensor array, the motion of a wireless vehicle can be manipulated by an airflow sensor array in a noncontact mode.As depicted in figure 3(g), a 2 × 2 array is built up with four sensors representing the front, back, left, and right directions, respectively.When the airflow blows any of the sensors in the array, the vehicle will move in the corresponding direction.

Intelligent healthcare
Flexible flow sensors play crucial roles in human healthcare for fluids monitoring, such as, superficial blood pressure induced by hemodynamics, blood flow direct detection within vessel, breath flow monitoring for respiratory disease management, etc.For instance, Lin et al proposed a fully integrated ultrasonic-system on patch for real-time monitoring of hemodynamics located in deep tissue [76], as illustrated in figure 5(a).Though relative movement between the device and the target tissue happened, the implanted machine learning algorithm (MLA) was able to select the best signal channel, and thereby allowed real-time monitoring of deep tissue hemodynamics even during human motion.Furthermore, the device standardized the data-interpretation process and expanded the accessibility of this powerful diagnostic tool in both inpatient and outpatient settings.
After vascular graft surgeries, the blood flow monitoring is crucial for patient recovery.Greatly different from traditional clinical wired implantable devices that requires careful fixation which easily causing infection, the booming developed flexible flow sensor have advantages in hemodynamics monitoring.For instance, Boutry et al reported a blood flow sensor consisted of a fringe-field capacitive sensor for pulsatile blood flow monitoring and a bilayer coil structure that was used for the transmission of radio-frequency data (figure 5(b)) [23].The device was made of biodegradable materials: poly(glycerol sebacate) with microstructured pyramids as a dielectric layer for the pressure sensitive regions, poly(octamethylene maleate (anhydride) citrate) and polyhydroxybutyrate/ polyhydroxyvalerate packaging layers, and poly(lactic acid) spacer used for the bilayer coils.The variations in capacitance resulted in a shift of the resonant frequency f 0 of the inductor-capacitor-resistor circuit, which was monitored wirelessly through the skin via inductive coupling with an external reader coil in a battery-free approach, the equivalent electrical circuit of which is shown in figure 5(b).It was validated in an artificial artery model and in a rat model, presenting advantages in real-time monitoring of hemodynamics after vascular surgery.Meanwhile, Herbert et al developed an intelligent stent integrated with wireless electronics system, which was capable of real-time detection of blood pressure, flow rate, and pulse rate (figure 5(c)) [82].The stent was fabricated by a laser-machining process achieving the balance between stent mechanics and wireless connectivity; the pressure sensors conformally integrated on which were prepared by fully aerosol jet-printing techniques.The sensing performance of device was both evaluated in a mimicking artery and in a rabbit model, proving a promising approach for continuous monitoring of hemodynamics to improve the treatment of vascular diseases.
As an unremitting behavior in our whole life, respiration involves inhaled/exhaled breathing cycles [83].The breathing biomechanics can be adapted as a significant physiological indicator for early detection of related illnesses.Fang et al developed a smart mask integrated with a breathing flow sensor for adaptive respiratory monitoring [84], as illustrated in figure 5(d).The flow sensor employed a permeable and anti-moisture textile as a triboelectric sensing layer, which was fabricated by Rayleigh instabilitiesassisted structured fibers.They also developed a deep-learning algorithm for respiration pattern classification with a high accuracy up to 100%.Furthermore, customized user-friendly wireless respiratory monitoring system was developed to connect users and clinicians, facilitating the development of personalized respiration management and intelligent healthcare.
The flexible flow sensors can also be employed to explore certain fluid behaviors within human organ, probably benefiting the treatment of corresponding diseases.For instance, Moshizi et al developed a miniaturized flow sensor, inspired by the vestibular hair cell sensors inside the human semicircular canals (figure 5(e)) [85].The bioinspired flow sensor was composed of vertically grown graphene nanosheets as sensing component, and penetrated PDMS as substrate.Embedded inside artificial lateral semicircular canal mounted on a rotary stage, the flow sensor was able to detect different frequencies and rotational angles in yaw, pitch, and roll.
Continuous detection of blood flow plays a crucial role in evaluating the status of vascular health for wide research and clinical scenarios.The fastly developed wearable and flexible flow sensors can improve the monitoring comfort level compared with traditional techniques, but they still face rigorous challenges, including limited accuracy and motion artifacts.To tackle these challenges, Webb et al presented a soft and skin-integrated ultrathin device for precise and continuous mapping of blood flow, based on thermal flow sensing mechanism (figure 5(f)) [86].The proposed device was composed of a single circular thermal actuator surrounded by two rings of sensors, designed for monitoring blood flow beneath a targeted area.The human subjects studies demonstrated that the device could perform precise assessment of both macrovascular and microvascular flow under series of physiological conditions, providing an effective approach for continuous monitoring during daily activities.

Intelligent city management
The phrase 'intelligent city' is becoming ubiquitous in the political, economic, scientific and colloquial discourse.Integrated sensor systems are a prerequisite for developing the concept of intelligent cities in practice because individual sensors can hardly meet the demands of the intelligent cities for complex information.In particular, flow sensors play an extremely important role in the construction of intelligent city management, such as air-conditioning network, smart monitoring of pipelines, water resource management, waterfall monitoring and so on.
Intelligent monitoring of water quantity in distribution networks and air flow in air-conditioning network is essential when talking about intelligent city management.A patchtype flexible flow sensor was developed to precisely control the supply air in large-scale air-conditioning network systems in buildings (figure 6(a)).The flow sensor was produced by applying photolithography onto a 25 µm thick polyimide film.The four sensors were attached at 90 • angles inside the surface of an 8 inches duct, and the obtained outputs were averaged [87].The relationship between the sensor output and the flow rate followed the King equation under the 0-3000 m 3 h −1 flow condition.Here, the King's law describes heat transfer from a cylinder of infinite length in terms of the resultingvoltage difference and is useful for hot-wire anemometry characterization.The constants are a complex combination of fluid thermal conductivity properties and flow geometry and should be found empirically [70]: where ∆V is flow induced voltage difference; v is velocity; a, b, c are constant coefficients.The proposed sensor enabled us to reduce fruitless energy consumption in these systems.
Riboldi et al presented a concept of a non-intrusive sensing node for smart monitoring of water with clamp-on sensors on pipes, piezoelectric transducers for ultrasound sensing, and copper electrodes for impedance (figure 6(b)) [88].The proposed sensing node could be easily clamped on plastic pipes to enable the measurement of multiple parameters without contact with the fluid, including water flow rate (up to 24 m 3 s −1 ), temperature, the pipe filling fraction, and ionic conductivity.The water resource is an essential field of economic growth, social progress, and environmental integrity.Ahmed et al offered a novel solution to meet water needs, distribution, and IoT-based quality management requirements [89], as shown in figure 6(c).The presented IoT-enabled Water Resource Management and Distribution Monitoring System (IWRM-DMS) employing various sensors, such as the water flow sensor, the pH sensor, the water pressure valve, the flow meters, and ultrasound sensors, to implement in rural cities.
The IWRM-DMS established the rural demand for water and the water supply system to minimize water demand.
Wolf et al proposed an intelligent swim pool equipped with an artificial LLS, a kind of flow sensors inspired by fish LLS, to assess a local hydrodynamic environment and further to classify the objects moving in water [24], as illustrated in figure 6(d).The artificial LLS was composed of eight alloptical 2D-sensitive flow sensors, for evaluating the hydrodynamic stimuli generated by different moving objects.With the assistance of machine learning, the detected hydrodynamic stimulus could classify objects with high accuracy of 98.6%.This validated that this passive flow sensing technique could be used for object detection, object identification, and collision avoidance in dark and murky environments.As shown in figure 6(e), Li et al develop an elastic, superhydro-phobic and conductive thin film enabled by a controllable composite of assembled CNT and elastomer for a smart umbrella [90].Through the adjustment of hydrophobic elastomeric coating, PDMS/CNTs/PDMS composite membrane (PCPM) with asymmetric three-layer structure was prepared.The surface wettability could be effectively controlled and still maintained superhydrophobic characteristics under the applied strain of 60%.The achieved film could function as a selfsupported smart umbrella to sensitively capture the simulated rain droplets with a series of frequencies and intensity.The self-supported smart umbrella could even response to the solar intensity for weather monitoring.More interestingly, owing the superhydrophobic feature, the umbrella acted as water rescue device to take the objects out of water and enabled a fast motion across the water surface.

Intelligent agriculture
Flexible flow sensors potentially facilitate the development of intelligent agriculture, improving agricultural practices and thereby increasing the production rate of crops.Various types of sensors can be deployed to collect physical and environmental data from the farms.Figure 7(a) presented the first flexible electronic sensing device that can harmlessly cohabitate with the plant and continuously monitor its stem sap flow, a critical plant physiological characteristic for analyzing plant health, water consumption, and nutrient distribution [25].Benefited from special design and materials selection, the realized plant-wearable sensor was thin, soft, lightweight and air/water/light-permeable, illustrating excellent biocompatibility, therefore enabling the sap flow detection in a continuous and non-destructive manner.The sensor was able to serve as a noninvasive, high-throughput, low-cost toolbox, and holds excellent potentials in phenotyping.
Wind, raindrop and water flow are particularly influential on fertilization, evaporation, and plant growth, and thus also affect yield.Flow sensors are widely used to support smart ploughing, sowing, irrigation, use of fertilizer and pesticides, harvesting, and livestock.Figure 7(b) exhibits a specific wind sensor that is used to support intelligent agriculture, which can be used for wind velocity detection for real-time monitoring and early warning of natural disasters [91].In addition to airflow sensors, water flow are also monitoring and used for intelligent agriculture.Figure 7(c) depicts prospective applications of the wind and water flow by scavenging environmental mechanical energy to provide distributed power supply for electrical devices in agricultural production [92].
Rapid iterations of sensing, energy, and communication technologies transform traditional agriculture into standardized, intensive, and smart modern agriculture.However, the energy supply challenge for the plentiful sensors or other microdevices constraints the extensive application of intelligent technologies in agriculture.By efficiently harvesting low-frequency mechanical energy from the  agricultural environment, including wind, rain, and water flow energy, TENGs can be a strong contender for distributed power for microdevice networks in intelligent agriculture [95].As shown in figure 7(d), to harvest the random lowfrequency vibration of the blade, a flexible single-electrode mode TENG based on PDMS film and laser-induced graphene preparation was proposed [93].The device could be attached to the blade with PDMS as the triboelectric negative layer and the blade as the triboelectric positive layer.The PDMS film achieved efficient breeze energy harvesting by contact separation with other blades under the action of wind through vibration.Inspired by the natural lotus leaf with hydrophobic interface, a TENG based on the single-electrode mode for raindrop energy harvesting in the agricultural environment was proposed, which consisted of a multioperating mode TENG and a micro-supercapacitor (figure 7(e)).The device was waterproof and self-cleaning, so the device could maintain stable output on wet and rainy days [94].Reprinted with permission from [99].Copyright (2016) American Chemical Society.

Other intelligent scenarios
As shown in figure 8(a), with the assistance of the desirable performance of GSEM-based airflow sensor, Zhou et al further demonstrated its potential application in a smart window system via flow velocity threshold control [81].People usually would like to open windows on windy and sunny days to keep indoor ventilation, and close windows on windy days to prevent bad weather from affecting indoor environment.The process of opening or closing windows could be intelligently controlled by GSEM-based airflow sensor, thus avoiding the trouble and inconvenience of human operation.Zhang et al introduced a nanonewton-scale biomimetic mechanosensor, and demonstrated the use of nanocrack-based electronic whisker (NCBEW) mechanosensor to sense the opening and closing of the door by monitoring the corresponding airflow [96], as shown in figure 8(b).When the door was opening or closing, the movement of the door panel induced positive or negative pressure, thus generating airflow disturbance.The experimental results validated that the ability of the device to track environmental changes via the perception of airflow.
With the rapid development of computers, microelectronics, and information processing technologies, higher requirements are put forward for machine to sense environmental information more accurately and quickly as a human.An et al mimicked the way that animals exploring the environments using hair-based sensors, designed a bendable biomimetic whisker mechanoreceptor (BWMR) for robotic tactile sensing [97], as illustrated in figure 8(c).Owing to the advantages of TENG technology, the BWMR converted external mechanical stimuli into electrical signals without a power supply, which was conducive to its widespread applications in robots.The sensor achieved a very high SNR with a minimum exciting force of 1.129 uN thanks to the simple and functional design, making it also suitable to detect natural flow vibration and object-induces air movement.
The use of renewable distributed energy sources such as wind, solar, and water wave to generate electricity has received worldwide attention.As shown in figure 8(d), Su et al proposed a proper large-scale system to harvest wind energy on highways with modular soft-contact anti-glare TENG (AG-TENG) devices in network [98].The system could be installed in the road median every certain distance according to the light blocking requirement.Therefore, the constant wind generated by the traffic flow was harvested by the AG-TENG system to realize close-looped self-powered sensing and identification for the smart and efficient highway.Conventional wind energy harvesting devices rely on electromagnetic induction and turbines, which are unsuitable for harvesting random, low frequency, and low-velocity wind energy due to their large size, high start-up wind speed, complex structural design, and high manufacturing cost.Compared with conventional wind energy harvesting devices, TENGs are more suitable for wind energy harvesting because of their versatile construction, lightweight, small size, and low start-up wind speed.As shown in figure 8(e), Zhao et al proposed a flag-type TENG for wind energy harvesting in any direction [99].The WTENG-flag was woven by conductive belts of Ni-coated polyester textiles (Nibelts) and Kapton film sandwiched Cu belts (KSC-belts).In each woven unit, a gap was leaven between the two electrodes so as to allow the contact-separation motion driven by wind fluttering.Under the action of wind energy, the Kapton film and the conductive cloth in the flag TENG come into contact and separate to realize the conversion of wind energy into electricity.Considering the lightweight, low cost, and scale-up facility, this WTENG-flag has great potentials for applications in weather/environmental sensing/monitoring systems.

Conclusion and prospect
Here, we present an overview of the recent developments of smart flexible flow sensors from the aspects of flow sensing technologies and related intelligent applications in our daily life.The representative application of various flexible flow sensors with different working principles and features in human intelligent life were summarized in table 1.The smart flexible flow sensors are stilled faced with many challenges, though they have rapid and booming developments.
First of all, the sensitivity, spatial distribution, and uniformity of smart flexible flow sensors require more attention, due to the demand of tiny flow dynamics variations monitoring and multiple monitoring sites.For the sensitivity improvement, we can learn from splendid flow sensing organs in nature, such as LLS in fish [16], cilia system in spiders [21].Their sensing mechanisms and unique structures can give us continual inspiration for the development of highly sensitive flexible flow sensors.Meanwhile, the effective distribution mode in these natural sensing systems provides us with a novel strategy to arrange spatially distributed flow sensors.Then, we can adopt Micro-electromechanical Systems (MEMS) or other MEMS-similar rigorous fabrication approaches to develop flexible flow sensors with high uniformity.
Secondly, multiple integration capability of smart flexible flow sensors should be improved.The smart flexible flow sensor system should be integrated with scenarionecessary functional modules, including conditioning circuits for weak signals amplification and signals processing, light and tiny batteries for powering the whole system, wireless modules for signals continuous transmitting, smart mobile terminals for data and results display, etc. forming a smart "human-flow field-interface" for users to friendly and conveniently monitor environments based on flow fields stimuli.
Thirdly, though there are several studies presenting the integration of smart flexible flow sensors with MLA for intelligent flow sensing, the development of intelligent flexible flow sensors/systems are confronted with challenges.The MLA can be used to optimize flexible sensors, from aspects of materials [100][101][102], structures [103], and distribution [22,104].Generally considering, MLA is always to process complicated and mass data, which is unable to be solved with traditional methods, such as linear calibrations and nonlinear fittings, etc.However, it is time-consuming, due to collection of larger data and rigorous data training process from flexible flow sensing systems.It requires to develop a smart MLA with simplified training process while maintaining high accuracy.
Fourthly, encapsulation is a crucial issue for flexible flow sensors to maintain performance stability, due to they have Zhao et al [99] to be exposed to fluidic environment for a long period.To work in the environments faced with chemical corrosion (for instance seawater), the flexible flow sensors are always to be coated a ultrathin encapsulation layer, such as PDMS, EcoFlex, Parylene C, etc.Among them, coating a parylene C layer is a foreground solution, due to the excellent chemical stability and the versatile processing capability (thickness at µm level) of this package material [105].On the other hand, we can transfer the shortcomings into advantages in specific scenarios, i.e. employing the environmental chemical information for flow field sensing.Poghossian et al developed an ion-sensitive field-effect transistor (ISFET)-based flow sensor, which was consisted of an ion generator and a pH ISFET that detected the in situ electrochemically generated H + or OH − ions [106].The experimental results revealed that the proposed flow sensor could precisely decode flow-velocity, flowdirection and diffusion-coefficient detection from flow fields.We envision that the smart flexible flow sensors will be comprehensively developed in the nearby future, with the unremitting efforts of researchers from different disciplines, including micro/nano fabrication, MEMS systems, wireless electronics, biomimetic engineering, artificial intelligence, IoT, etc.It has promising applications in our daily life from various aspects of intelligent navigation, intelligent healthcare, intelligent city management, intelligent agriculture, intelligent energy collection, etc.

Figure 2 .
Figure 2. Typical natural flow sensing systems.(a) Fish lateral line system for hydrodynamic sensing.(i) Neuromasts distribution on fish body, (ii) illustration of working mechanism in neuromast, iii) two main types of neuromasts.Reproduced from [16].© IOP Publishing Ltd All rights reserved.(b) Harbor seal whiskers for hydrodynamic sensing.(i) Harbor seals, (ii) cross-section of the harbor seals whiskers, (iii) microstructure of seal whisker, (iv) CFD simulation results of the wakes generated by different whiskers.Reproduced from [18].© IOP Publishing Ltd All rights reserved.Reproduced with permission from [19].[insert copyright line, if specified].(c) Sensing systems of spiders and crickets for aerohydrodynamic sensing.(i) Trichobothria appeared on the spider's leg, (ii) filiform hairs presented on the cricket's cerci, (iii) cuticular socket, (iv) schematic diagram of the flow-sensing hairs.Reproduced from [20].© IOP Publishing Ltd All rights reserved.

Figure 3 .
Figure 3. Schematic illustrations of engineering strategies for flexible flow sensing.(a) Typical configurations for mechanical deformation-based flexible flow sensors.(i) Solid film, (ii) diaphragm, (iii) cantilever, (iv) cilium.(b) Schematic illustrations of transduction mechanisms in flexible flow sensors based on mechanical deformation strategy.(i) Piezoresistive mechanism converts displacement d into resistance R variation, (ii) a capacitive mechanism converts displacement d into capacitive C variation, (iii) optical mechanism converts displacement d into gratings displacement x 0 variation, (iv) piezoelectric mechanism converts displacement d into charge q variation, (v) triboelectric mechanism converts displacement d into charge q variation, and (vi) magnetoelastic mechanism converts displacement d into magnetic flux density B variation.(c) Schematic illustrations of transduction mechanisms in flexible flow sensors based on thermal type.(d) Schematic illustrations of transduction mechanisms in flexible flow sensors based on ultrasound technique.

Figure 4 .
Figure 4. Flexible flow sensor-assisted intelligent navigation.(a) Fish constricted canal lateral line system inspired a flexible flow sensor for underwater obstacles recognition.Reproduced from [77], with permission from Springer Nature.(b) Fish bimodal sensing lateral line system inspired a pressure/velocity sensing system mounted on UUV model for underwater obstacles location in 3D space.© [2022] IEEE.Reprinted, with permission, from [22].(c) Fish lateral line system inspired flow sensing-based control methodology benefiting UUV's precise navigation.Reproduced from [78], with permission from Springer Nature.(d) A novel flexible airflow sensor based on four curved microcantilevers arranged in a cross-form configuration for flow field monitoring and angle of attack (α) detection.Reproduced from [21], with permission from Springer Nature.(e) A micro hot-film flow sensor array for aerodynamic parameters detection.Reproduced from [79].© IOP Publishing Ltd All rights reserved.(f) An artificial hair sensor (AHS) mimicking those used by natural fliers (such as bats, locusts, and crickets) to detect surface flow features.Reproduced from [80].© IOP Publishing Ltd All rights reserved.(g) A wireless vehicle can be manipulated by an airflow sensor array in a noncontact mode.[81] John Wiley & Sons.© 2021 Wiley-VCH GmbH.

Figure 5 .
Figure 5. Flexible flow sensor-assisted intelligent healthcare.(a) A skin-integrated ultrasound array for CVD assessment.Reproduced from [76], with permission from Springer Nature.(b) An artery-integrated blood flow sensor for postoperative healthcare.Reproduced from [23], with permission from Springer Nature.(c) A 3D-printed all-in-one stent sensor for hemodynamics monitoring.Reproduced from [82].CC BY 4.0.(d) A deep-learning-assisted mask-integrated flow sensor for adaptive respiratory monitoring.[84] John Wiley & Sons.© 2022 Wiley-VCH GmbH.(e) A bioinspired flow sensor with porous graphene for monitoring of physiological movements within lateral semicircular canal.Reproduced from [85].CC BY 4.0.(f) An ultrathin and skin-conforming sensor technology for noninvasive and continuous mapping of blood flow.Reproduced from [86].CC BY 4.0.

Figure 6 .
Figure 6.Flexible flow sensor-assisted intelligent city management.(a) A patch-type flexible flow sensor for precise control the supply air in large-scale air-conditioning network systems in buildings.Reprinted from [87], Copyright (Year), with permission from Elsevier.(b) A non-intrusive sensing node for smart monitoring of water with clamp-on sensors on pipes.Reproduced from [88].CC BY 4.0.(c) A novel solution to meet water needs, distribution, and IoT-based quality management requirements.Reprinted from [89], Copyright (Year), with permission from Elsevier.(d) An intelligent pool-integrated with flexible flow sensor for obstacles recognition.Reproduced from [24].CC BY 4.0.(e) A Self-supporting smart umbrella for rainfall detection.Reproduced from [90].CC BY 4.0.

Figure 7 .
Figure 7. Flexible flow sensor-assisted intelligent agriculture.(a) A water flow sensor used to measure the physiological characteristics of an individual plant.Reproduced from [25].CC BY 4.0.(b) An energy harvesting and sensing device based on electromagnetic-triboelectric hybrid generator.Reprinted with permission from [91].Copyright (2021) American Chemical Society.(c) A self-powered automatic irrigation, weather monitoring and wireless water level warning multifunctional management systems.Reprinted from [92], Copyright (2019), with permission from Elsevier.(d) A flexible single-electrode mode TENG based on PDMS film and laser-induced graphene for energy harvesting by contact separation with other blades under the action of wind through vibration.[93] John Wiley & Sons.© 2021 Wiley-VCH GmbH.(e) A TENG based on the single-electrode mode for raindrop energy harvesting in the agricultural environment.Reprinted from [94], Copyright (2020), with permission from Elsevier.

Figure 8 .
Figure 8. Flexible flow sensors for other intelligent scenarios potential application in intelligent and digital life.(a) A GSEM-based airflow sensor for application in a smart window system via flow velocity threshold control.[81].John Wiley & Sons.2021].(b) The use of NCBEW mechanosensor to sense the opening and closing of the door by monitoring the corresponding airflow.Reproduced from [96].CC BY 4.0.(c) A bendable biomimetic whisker mechanoreceptor (BWMR) is designed for robotic tactile sensing.[97] John Wiley & Sons.© 2021 Wiley-VCH GmbH.(d) The proper large-scale system to harvest wind energy on highways with modular soft-contact AG-TENG devices in network.[98] John Wiley & Sons.© 2023 Wiley-VCH GmbH.(e) Flag-type TENG for wind energy harvesting in any direction.Reprinted with permission from [99].Copyright (2016) American Chemical Society.

Table 1 .
Summary of characteristics of flexible flow sensors for representative intelligent application scenarios.