Proposal of a Method for Detecting Sensor Failure Using General-Purpose Sensing Circuit

In recent years, home appliances and other devices have become connected to networks, increasing the demand for IoT devices, and thus the demand for sensors to be installed in them. There are various types of sensors, and it is necessary to prepare detection circuits for each type of sensor. In addition, there are many types of sensors, and it is necessary to prepare a detection circuit for each type of sensor. Preparing detection circuits for each type of sensor increases the number of devices and makes them expensive. In some cases, devices are purchased from a specific manufacturer and manufactured only after receiving an order, so it takes a long time to repair or replace them. In addition, it is necessary for a person to judge the necessity of repair. To solve these problems, we propose a method for detecting sensor failures and degradation using a general-purpose sensing circuit, as well as a calibration method for degraded sensors. In this paper, we propose a method of detecting sensor failure and degradation and a method of calibrating degraded sensors using a general-purpose sensing circuit, so that failed sensors can be automatically replaced and degraded sensors can be calibrated.


Introduction
In recent years, home appliances and other devices have become connected to networks, increasing the demand for IoT devices.The Fig. 1 shows the trend of the number of IoT devices in the world (1) .From this figure, we can see that the number of IoT devices continues to increase.Most Fig. 1.Global IoT device count trends and forecasts (1) . of the IoT devices are equipped with sensors, and the demand for sensors is expected to increase at the same time.There are various types of sensors, and it is necessary to prepare a detection circuit for each type of sensor.In addition, it is necessary to prepare a detection circuit for each type of sensor, which requires a large number of devices and is expensive.In addition, when a failure occurs, the device is purchased from a specific manufacturer and manufactured only after receiving an order, which takes a great deal of time.
In order to solve these problems, a general-purpose sensing system that can measure multiple types of sensors with a single sensing circuit has been proposed in previous studies (2)(3)(4) .With this system, different types of sensors can be repaired or replaced by simply replacing the sensor element in the input terminal of the sensing circuit.However, the system in the previous studies does not have a system for detecting sensor degradation or failure, so it is necessary for a person to judge the necessity of replacing the sensor.In this system, the sensor element can be replaced simply by replacing it, so if the system can detect a sensor failure, it can easily be replaced automatically.In addition, if the system can detect the deterioration of a sensor element, it can correct the measured value of the sensor.If the degradation of the sensor element can be detected, the measured value of the sensor can be corrected.If the sensor can be corrected, the life of the sensor can be extended and the cost can be reduced.
In this study, we propose a method to detect sensor failure and degradation using a circuit proposed in previous studies.3 sensors are connected, and sensor failure and degradation are detected by minority vote.We assume that the three sensors have the same characteristics, are installed in the same environment, and that two of the three sensors are normal.In this study, we assume that the three sensors have the same characteristics, are installed in the same environment, and that two of the three sensors are normal.This simplifies the replacement of the sensors and is expected to increase the service life and reduce the cost of the sensors.

Previous study 2.1 Method to identify energy conversion type and energy control type
In our previous studies, we defined that sensors can be classified into five types based on the type of output electricity: voltage type, current type, resistance type, capacitance type, and inductance type.In order to adapt to these five types of sensors, they classify them into energy conversion type sensors (voltage type, current type) and energy control type sensors (resistance type, capacitance type, inductance type), and identify the sensors connected to the circuit as one of them (2)(3)(4) .An energy conversion type sensor is a sensor that converts input energy such as light or heat input to the sensor into electrical energy and outputs it.On the other hand, energy control type sensors change the impedance of the sensor itself depending on the input energy such as heat or light.The energy control type sensor changes its impedance depending on the input energy such as heat or light.The supplied energy is controlled by changing the impedance of the sensor itself, and the energy is output as electrical energy (5) .In our previous research, we identified various types of sensors as either energy conversion type or energy control type, and switched the configuration of the detection circuit to enable measurement of various types of sensors.

Measurement method
Fig. 2 shows a detection circuit from a previous study (2)(3)(4) .This circuit is based on an inverting amplifier circuit, and terminal B is the input terminal to which the sensor is connected.Switch A is switched according to the type of sensor, and the power supply is connected to GND[V] when an energy conversion type sensor is connected, and the power supply is connected to AC power supply VAC when an energy control type sensor is connected.This enables sensor measurement.
Eqs. 1 -5 are the input-output relation equations in the circuit shown in Fig. 2. From these equations, it can be confirmed that the output voltage VOUT changes with respect to the sensor voltage VIN, current IIN, resistance RIN, capacitance CIN, and inductance LIN.The voltage-type and current-type sensors are energy conversion sensors, and the output voltage VOUT is used as the output of the sensing circuit.Resistance, capacitance, and inductance sensors are energy control type sensors, and the output voltage VOUT is an AC voltage.Therefore, when measuring energy control type sensors, the rectified and smoothed voltage VFLAT of the output voltage VOUT is used as the output of the detection circuit.

Proposed method 3.1 Sensor failure and degradation detection methods
This section describes the detection method of sensor failure and degradation in this study.First, we obtain the measured values of three sensors in the system of the previous study.The output of the failed sensor does not change and GND[V] is output.Therefore, when the measured values of the two normal sensors are changing, if the measured value of the failed sensor remains constant at GND[V], the failure can be detected.Table 1 shows the conditions for determining degradation detection based on the measured values of each sensor.The measured values of the three sensors are S1, S2, and S3.The difference between the measured values of sensor 1 and sensor 2 is S21, and the difference between the measured values of sensor 1 and sensor 3 is S31.
(a) All three sensors are normal The sensor measurements S1, S2, and S3 are all the same value.Therefore, if the difference between sensor 1 and sensor 2 is S21 = 0 and the difference between sensor 1 and sensor 3 is S31 = 0, then all the values are the same and the sensor is normal.
(b) Sensor 1 is degraded Sensor 2 and Sensor 3 are normal sensors, and the measured values of S2 and S3 are the same.Because sensor 1 is degraded, there is a large difference between the measured value of sensor 1 and that of the normal sensors.Therefore, if S21, S31 ≫ 0 and S21 = S31 , it can be judged that sensor 1 is deteriorated.

(c)
Sensor 2 is degraded Sensor 1 and Sensor 3 are normal sensors, and the measured value of Sensor 1, S1, and the measured value of Sensor 3, S3, are the same.Because sensor 2 is degraded, there is a large difference between the measured value of sensor 2 and the measured value of the normal sensor.Therefore, if S21 ≫ 0 and S31 = 0, it can be judged that sensor 2 is deteriorated.
(d) Sensor 3 is degraded Sensor 1 and Sensor 2 are normal sensors, and the measured value of Sensor 1, S1, and the measured value of Sensor 2, S2, are the same.Because sensor 3 is degraded, the measured value of sensor 3, S3, is significantly different from the measured value of the normal sensor.Therefore, if S21 = 0 and S31 ≫ 0, it can be judged that sensor 3 is deteriorated.

The Method of Calibrating Degraded Sensors
Calibration of degraded sensors is performed.This section describes the calibration method.Fig. 3 shows the flow of deriving the correction equation. (a) As shown in the graph in Fig. 4, the measurement of a normal sensor and the measurement of a degraded sensor are performed while changing the input of physical quantities such as heat and light.The measured value of the normal sensor is the true value T, and the measured value of the degraded sensor is the degraded value D.

(b)
The relationship between the true value T and the degraded value D is expressed as an nth-order polynomial, as shown in Eq. 6, and the coefficient a of each term is determined.The degree of this polynomial should be large Table .1. Judgment conditions for degradation detection.enough so that there is no error in the degraded value relative to the normal value.This polynomial is used as a correction equation.Once the correction equation is derived, the degraded sensor can be used as a normal sensor by substituting the degraded value D in the calculation.

Verification experiment 4.1 Experimental method
We constructed an experimental circuit based on Fig. 2, and connected a circuit consisting of three sensors and a switch to terminal B as shown in Fig. 6.By installing and controlling the switch, the three sensors can be switched for measurement.In this experiment, we will use a CdS cell as an example.As shown in Fig. 7, three CdS cells were connected in the same position as much as possible, and the sensors S1, S2, and S3 were from left to right.In this experiment, the surface of the  light-receiving part of sensor S1 was scratched to make it a degraded sensor.Under these conditions, we verified whether it was possible to detect a degraded sensor and calibrate the detected sensor.

Result
The results of the sensor degradation detection experiment are shown in Table 2.The measured values of S2 and S3 of the normal sensor are almost the same.On the other hand, there is a large difference in the measured value of S1 of the degraded sensor compared to that of the normal sensor.Therefore, S1 is judged to be "degradation".
Fig. 8 shows the results of the calibration of the degraded sensor.the output T of the normal sensor and the output D of the degraded sensor were measured while changing the amount of light received by the CdS cell.The blue plots were connected by lines, and a fourth-order polynomial was derived using spreadsheet software.By substituting the measured value D of the degraded sensor into the polynomial, we obtained the red plot in Fig. 8. Therefore, by correcting the measured value D of the degraded sensor, it became possible to use the degraded sensor as well as the normal sensor.As a result, we confirmed that it is possible to detect a degraded sensor and to correct the measured value of the degraded sensor by a minority vote using three sensors.Therefore, even if a sensor deteriorates, it can continue to be used, and a long life of the sensor can be expected.

Conclusions
In previous studies, a general-purpose sensing circuit that is adaptable to five types of sensors has been proposed to simplify the replacement of failed sensors and to reduce the cost.If this sensing circuit can detect sensor failures and degradation, failed sensors can be replaced automatically, and degraded sensors can continue to be used because their measured values can be corrected.Therefore, it is expected to simplify the replacement of sensors, increase the service life, and reduce costs.
In this study, we proposed a method for detecting sensor failures and degradation and a method for calibrating degraded sensors using the circuit proposed in the previous studies.3 sensors are used, and assuming an environment in which only one of the three sensors degrades, it is possible to detect and calibrate the degraded sensor by minority voting.The verification showed that it is possible to detect and calibrate degraded sensors.Therefore, even if a sensor deteriorates, it can continue to be used, and a long life of the sensor can be expected.