Cold starting temperature time-related compensation model of inertial sensors based on particle swarm optimization algorithm

Abstract

With the miniaturization of inertial instruments, sensors mounted inside are vulnerable to interference. In a complex thermal transmission environment, temperature drift is the main factor restricting the precision of high-performance inertial sensors. To solve this problem, a new method for compensating the time-related cold starting temperature drift of the inertial sensors is introduced in this paper. Based on the perspective that temperature drift can be regarded as the response curve of the sensor system to temperature and temperature gradient, temperature compensation models of first-order, second-order, and higher-order are proposed. Meanwhile, the particle swarm optimization algorithm is used to solve the model parameters. Under various practical circumstances, the method can be used to flexibly compensate the temperature drift and reduce the standard deviation of the output signal by about four times. Compared to other models or algorithms, the simulation and experimental results indicate that the proposed model is superior in adaptability, stability, and reliability.

Cold Starting Temperature Drift Modeling and Compensation of Micro-Accelerometer Based on High-Order Fourier Transform

“…Specifically, the stability of the cold start-up output can be maintained at about 3 × 10 −5 g under different ture conditions, and the maximum drift error can be kept within 5 × 10 −5 g theoretically, which are good enough measures to meet the requirements of a fast and stable start for a high-precision inertial navigation system. Additionally, after practical verification, the proposed temperature compensation model based on high-order Fourier transform for the cold start-up phase has the following advantages over the time-related compensation model proposed in the literature [19]:…”

“…Specifically, the stability of the cold start-up output can be maintained at about 3 × g under different temperature conditions, and the maximum drift error can be kept within 5 × g theoretically, which are good enough measures to meet the requirements of a fast and stable start for a high-precision inertial navigation system. Additionally, after practical verification, the proposed temperature compensation model based on high-order Fourier transform for the cold start-up phase has the following advantages over the time-related compensation model proposed in the literature [ 19 ]: The proposed model possesses a simple structure, and its basic expression can be described as the accumulation of trigonometric functions, which makes it easier to be programmed in the processor than the time-related temperature compensation model for containing exponential terms; The time-related model contains time variables and exponential functions so that the resource consumption in the processor gradually increases with time. The proposed model is only related to temperature when model parameters are determined, which greatly increases the engineering practicability of the model; The particle swarm optimization (PSO) used in the time-related model has poor convergence ability in the case of multivariate large-scale optimization, so as a result, it has large residuals of model fitting and low efficiency for model identification.…”

“…To verify the feasibility and efficiency of the proposed temperature compensation model, data for the repeatability test were utilized to identify the time-related model proposed in [ 19 ] and the Fourier-related model for the cold start-up phase. After that, we program the model parameters and set the processor to output the data before and after temperature compensation with two different models.…”

“…It can be seen from Table 1 that the output characteristics of the micro-accelerometer in the cold start-up phase are improved after temperature compensation with two different models. Nevertheless, compared with the time-related model proposed in [ 19 ], the temperature compensation model on the basis of high-order Fourier transform has a better effect on compensating the output drift of the accelerometer in the cold starting section. After compensation with the time-related model, since the time characteristics of the output curve in the cold start-up phase are not exactly same as those of the impulse response curve, the compensation for the initial rapid rising section is so poor that there is only slight improvement in the stability and the maximum drift error compared with the original data.…”

“…However, the verification was tested for about 40 h; thus, it is difficult to research the characteristics in the start-up phase. Corresponding to this, reference [ 19 ] creatively proposed a time-related temperature compensation model specifically for the cold starting output of the accelerometer. This model approximates the complex output in the cold start-up phase to the response curve related to time and temperature and then utilizes the particle swarm optimization (PSO) algorithm to identify the model parameters.…”