Is too much IoT sensor really good?

Recently, I found that the car's "check engine" warning light is on, which may indicate a minor problem with the engine, which may cause the car to overheat. I quickly checked with the on-board diagnostic code scanner and found that the warning light was on in some respects related to the coolant of the engine. However, the temperature indicator of the driving console (dashboard) still shows that the engine temperature is "normal". Later, I discovered that the problem is not related to the powertrain, but the sensor failure of the cooling system. For me, modern super-induction vehicles are well-known IoT pioneers. In fact, for many models, this is an IoT phenomenon because the car can be connected to the manufacturing plant to report various readings, status and events. Whether the car is directly connected as an IoT node, today's vehicles are equipped with a number of sensors that have a self-awareness of various readings such as temperature, pressure, flow and on/off. We are also told that it is really wonderful to have it all.

But I am not sure if this is the case. Any experienced engineer knows that the sensor is the most vulnerable part of the system. Due to their inherent role, sensors must often be exposed to the real world of moisture, vibration, temperature and other physical stresses. Sometimes such environmental exposure is directly caused by the targets it monitors, but often comes from side effects that occur when monitoring other parameters. Regardless of the cause, the life of the sensor is more stringent than the electronic components on a typical circuit board (PCB), even if it is compared to board components in a vehicle environment. The problem is that when we add IoT sensors to everything, we will see more false positives and false negatives, making testing and evaluation more and more difficult. Soon, planning how to test the authenticity and credibility of many readings and warning indicators will be a big part of the test. Although this is a problem, it is only part of it. It is quite difficult to evaluate the sensor and test its readings. Because the options are very limited or just because they are not attractive. You can add redundant sensors and interface circuitry, but this adds cost, weight, power, and space burden, and you need a way to determine which of the two sensors is correct. Or, you might have to add two extra sensors and then use the two-out-of-three scheme? Of course, all additional added redundancy may also increase the reliability of the signal chain associated with the sensor. Another option is to build true, independent and closed circuit tests for sensor performance. In some cases, this is a very practical approach. For example, you can direct the motor to a specified position or speed and then see if the sensor reading is consistent with the orientation action. If so, the starter and sensor are likely to be good; if they are inconsistent, there must be something wrong and must be checked further. However, it is apparent that this stimulation/response phenomenon is completely impractical for many sensor variables or settings, such as temperature or pressure readings. I am thinking that the expansion and popularity of IoT-based sensor nodes will have several unintended consequences. First, when too many alarms become a problem, there is a tendency to ignore many alarms. Although this may not be a good response, it is a normal human response to the constant occurrence of such stimuli, especially when many of them are considered false alarms. Second, test engineers will have to spend more time designing algorithms that correlate multiple sensor readings to determine if they can figure out better conclusions—such as which readings are correct and which are wrong. In applications that use multiple sensors, it is also common to have readings that are out of range but still related to other readings. For example, overheating (temperature readings) may be related to insufficient coolant level or flow indication. But these associations are not easy to build, and require a lot of simulation, modeling, and especially system-level understanding, and when subtle interactions and relationships increase the complexity of the system, it becomes increasingly difficult to achieve. . Have you ever encountered too many sensor data like IoT causing too many false alarms? And the unnecessary system shutdown that comes with it? Or ignore all alerts directly? Are you worried that too many sensors can cause eventually uncontrollable? Also, will these sensor readings put a burden on test development?