多传感器目标检测的性能评估（一）

通过对传感器性能分析,将ROC曲线予以量化,根据其几何意义导出传感器探测性能的评估模型,该模型是一个单一的量化指标,用观测量的概率均值和方差来评估传感器性能的优劣。     

Detection Performance of a Mean-Level Threshold

The problem of detecting signals in nonstationary clutter is met by presenting a mean-level or adaptive threshold which adjusts to the changing background level. Such a threshold performs better than a fixed threshold that must be set for the highest amplitude clutter. However, the mean-level threshold does not perform as well for stationary noise as a fixed threshold set at the proper value. One measure of effectiveness of an adaptive threshold is its performance in stationary noise (compared to the optimum fixed threshold) for a specified speed of response. For the mean-level threshold, a simple mathematical solution is found for the detection probability when the noise is stationary and the signal scintillates rapidly. The performance is evaluated for a wide range of mean-level-threshold time constants and for several false-alarm probabilities. The results are presented graphically. As an example, the mean-level threshold suffers 3 dB in detectability (equivalent signal-to-noise ratio) in the presence of stationary noise as compared to the optimum fixed threshold for 50-percent probability of detection, false-alarm probability of 10-8, and an adjustment time of 15 times the signal duration.     

Detection Performance of Soft-Limited Phase-Coded Signals

General analytic expressions are developed for the soft-limited digital pulse compressor (matched filter). This theoretical development is then used for the hardware realization of a two-channel (I,Q), 3-bit-limited digital pulse compressor with a compression ratio of 255: 1. The realized hardware uses state of the art integrated circuit devices. An experimental laboratory setup is described. This setup is used to study hard-limited versus 3-bit-limited matched-filter performance characteristics with the data in the following areas: 1) constant false alarm rate (CFAR) characteristics as a function of threshold settings and noise levels; 2) single target detection characteristics as a function of input signal-to-noise ratio (SNR); and 3) two target performance characteristics: a) the amplitude of a weaker target as a function of target ratio and target overlap; and b) the detection characteristics of a weaker target as a function of weaker target SNR, strong target SNR, and target overlap.     

Detection Performance of Hard-Limited Phase-Coded Signals

The performance of a complex phase-coded waveform digital processor with hard-limiting constant false-alarm rate (CFAR) is presented. Processing losses relative to ideal matched-filter performance are computed and verified by hardware measurement. The losses considered include the consequences of hard limiting, envelope algorithm implementation, range cusping, and the associated effects of code length, IF filter bandwidth, and in-phase and quadrature channel phase offset. The range resolution properties of two closely spaced targets are also considered.     

Exact Detection Performance for Broadband Correlators

The detection of a broadband plane wave in an isotropic noise field is considered. The detection statistic used is the cross correlation of the sampled output from two sensors where the sampling rate equals the Nyquist rate and the time delay is an integer multiple of the sampling period. By restricting the number of points in the cross correlation to be even, the detection and false-alarm probbilities are obtained in closed form. Numerical results are presented for several special cases.     

Digital Control of High Performance Aircraft Using Adaptive Estimation Techniques

Since the early 1960's, a rapid advance in signal processing, including filtering and estimation techniques, has been evident. In contrast, applied feedback control, particularly for aircraft, is currently based on technology available prior to 1960, i. e., primarily either constant gain feedback or at most a standard gain-scheduling. In this paper, an adaptive signal processing algorithm is joined with gain-scheduling to produce an effective scheme for controlling the dynamics of high performance aircraft. A technique is presented for a reduced-order model (the longitudinal dynamics) of a high performance short-takeoff-and-landing (STOL) aircraft. The actual controller views the nonlinear behavior of the aircraft as being equivalent to a randomly switching sequence of linear models taken from a preliminary piecewise-linear fit of the system nonlinearities. The adaptive nature of the estimator is necessary to select the proper sequence of linear models along the flight trajectory. From the analysis of the reduced-order model the nonlinear behavior has been found to be well approximated by assuming an effective switching of the linear models at random times, the durations of which reflect the motion of the aircraft in response to pilot commands.     

高性能叶轮机全3维叶片技术趋势展望

高性能叶轮机是驱动先进航空发动机/地面燃气轮机发展的核心技术,为明晰叶轮机技术发展问题与趋势,在相关文献调研基础上,从基元叶栅、展向积叠、端区处理、精细化设计以及全局观念等方面出发,概要分析阐述了中国叶轮机全3维叶片技术继续发展的要点与突破口,指出全面综合最大折转亚声速叶栅、允许分离超声速叶栅、弱化激波叶栅、掠弯参数化积叠、叶身/端壁融合等基础研究成果并结合伴随方法进行精细化设计的负荷最大化技术,再辅以3维空间、非定常流动、细节关联等全局观念下派生的技术是全3维叶片技术的重要发展方向.未来全3维叶片将注重全3维空间流线曲率的良好控制.     

燃烧室火焰筒头部性能对比试验

根据前期某型发动机燃烧室头部降低冒烟数性能试验结果,火焰筒头部结构存在高温烧蚀变形痕迹,为保证该结构工 作可靠性和使用寿命,将原型燃烧室火焰筒头部传统的孔板+挡溅板的冷却结构改进为带有一定角度的收敛双锥形冷却结构,并 在单管燃烧试验器、扇形燃烧室试验件上采用连续气源、模拟参数进行了性能对比试验,录取了改进前后2种头部的点熄火边界、 燃烧效率及头部壁温。结果表明：在同工况下,改进后的火焰筒头部着火余气系数更大,点火边界更宽;贫油熄火边界相当;燃烧 效率基本相当,均大于0.99,符合性能要求;改进后火焰筒头部壁温较原型的有较大降低,温度分布更均匀。     

Design and Performance Comparison of Distributed Detection Networks

Optimization of a distributed detection network using theminimum global cost criterion results in local processors thatindividually form the likelihood ratio when the input observationvectors are statistically independent. In addition, the localthresholds and the network performance can be expressed as afunction of the receiver operating characteristics (ROCs) of the localprocessors. The performance of rive distributed networks arecompared numerically using local ROCs from the conic ROCfamily.     

Detection Performance of Logarithmic Receivers Employing Video Integrators

The detection performance of logarithmic receivers in Rayleigh and non-Gaussian clutter is investigated. In Rayleigh clutter the performance is determined for steady, Swerling case 1, and Swerling case 2 targets. The detection loss of logarithmic receivers is generally less than the ? log n loss conjectured by Green, but consistent with the 1.08-dB asymptotic loss established by Hansen. The Swerling case 2 loss, important in frequency- agility applications, canbe severe for a small number of integrated pulses and high Pd, and apparently approaches the 1.08-dB asymptotic loss as a lower bound. Graphs of GramCharlier series cumulants are provided to allow determination of logarithmic-receiver performance. Curves are presented to allow the detection performance of logarithmic receivers in log-normal and Weibull clutter to be determineds.     

Detection Performance in Clutter with Variable Resolution

Experiments were conducted to determine the influence of background clutter on target detection criteria. The experiment consisted of placing observers in front of displayed images on a TV monitor. Observer ability to detect military targets embedded in simulated natural and manmade background clutter was measured when there was unlimited viewing time. Results were described in terms of detection probability versus target resolution for various signal to clutter ratios (SCR). The experiments were preceded by a search for a meaningful clutter definition. The selected definition was a statistical measure computed by averaging the standard deviation of contiguous scene cells over the whole scene. The cell size was comparable to the target size. Observer test results confirmed the expectation that the resolution required for a given detection probability was a continuum function of the clutter level. At the lower SCRs the resolution required for a high probability of detection was near 6 line pairs per target (LP/TGT), while at the higher SCRs it was found that a resolution of less than 0.25 LP/TGT would yield a high probability of detection. These results are expected to aid in target acquisition performance modeling and to lead to improved specifications for imaging automatic target screeners.     

Optimal Detection and Performance of Distributed Sensor Systems

Global optimization of a distributed sensor detection system withfusion is considered, where the fusion rule and local detectors aresolved to obtain overall optimal performance. This yields coupledequations for the local detectors and the fusion center.The detection performance of the distributed system with fusionis developed. The globally optimal system performance is comparedwith two suboptimal systems. Receiver operating characteristics(ROCs) are computed numerically for the problem of detecting aknown signal embedded in non-Gaussian noise.     

Performance Data for a Double-Threshold Detection Radar

A problem f requently encountered by radar systems analysts is the evaluation of the "double-threshold" or M out of N detection process. Detection probabilities of this process are binomially distributed, making it difficult to obtain exact results for large values of the number of samples and for low probabilities of false alarm. In this paper, the M out of N detection algorithm is defined and detection performance is calculated for the special cases of the nonfluctuating target and Swerling cases I and 11 for false alarm probabilities of 106, 10-8 and 10-10.     

The Detection Performance of a Modified Moving Window Detector

The modified moving window detector for binary integration has a shorter window length compared with the ordinary moving window detector because of the addition of a third threshold. When the second threshold is reached, a counter begins to count until the third threshold is reached. A graphical method using a probability chain is developed here to analyze and to calculate the detection performance of this type detector. The detection performances for some practical cases are calculated by computer and are compared with the ordinary moving window detector.     

雷达检测性能试验等效推算方法

 雷达检测性能是雷达最基本、最重要的性能,雷达检测性能试验是雷达试验的重要内容。由于各种原因,针对雷达的系统级原型试验常常难以开展或者只能极少量地开展,通常只能通过进行大量等效试验来辅助雷达性能的鉴定评估工作。本文针对缺乏原型试验条件下的雷达检测性能评估问题,提出了一种基于试验数据分析的雷达检测性能等效推算方法。首先给出了等效试验与原型试验之间的差异性度量方法,它是等效推算和性能评估的基础;接着对存在原型试验条件下的等效推算方法进行了研究;而后针对通常无法进行原型试验的实际情况,提出了一种基于预测模型和仿真试验相结合的零子样原型试验条件下的等效推算方法;最后通过仿真实例验证了本文方法的合理性和可行性。     

Limiter-Discriminator Detection Performance of Manchester and NRZ Coded FSK

The theory of limiter-discriminator detection of a frequency shift keyed (FSK) carrier is reviewed and this theory is used to predict the bit error probability performance of Manchester coded and nonreturn-to-zero (NRZ) coded data. A major result of the study is that the predetection time bandwidth product BT and the devia tion ratio h needed to give optimum performance for Manchester coding are seen to be larger than optimum NRZ FSK. Specifically, BT ~ 2 and h ~ 1 will result in Manchester performance about 2 dB worse than optimum NRZ.     

基于ResNet-LSTM的航空发动机性能异常检测方法

为了实现数据驱动的航空发动机性能异常的智能检测,提出了一种基于残差网络（ResNet）-长短期记忆网络（LSTM）的发动机性能异常检测方法。采用发动机性能数据图像化方法,在数据降维的同时,完备保留数据的关联特征和时序特征;以残差单元构建发动机性能异常检测模型,在加深网络结构的同时,消除深层网络梯度消失问题,提高发动机性能图像空间关联特征的提取能力。同时,引入LSTM,提出基于ResNet-LSTM的发动机性能异常检测模型,通过ResNet与LSTM的融合,强化异常检测模型对时序特征的提取,提升发动机性能异常检测的准确率;通过发动机运行数据进行验证。结果表明：在训练集上,该方法的异常检测准确率为94.95%,比基于ResNet18、ResNet34、ResNet50异常检测模型的分别提高10.87%、8.00%、3.23%;在测试集上,该方法的异常检测准确率为92.15%,比基于ResNet18、ResNet34、ResNet50异常检测模型的分别提高11.81%、9.45%、3.78%。     

船载雷达速率陀螺动态性能检测方法

提出了利用同船装载的惯性导航系统船摇数据检测雷达速率陀螺动态性能的新方法。通过理论分析,建立了利用惯导船摇数据检测雷达速率陀螺性能的数学模型,设计了检测方法和数据处理方法,并进行了试验验证及其检测精度分析。     

钛合金胶接点焊接头的性能检测

研究了不同焊接电流下钛合金板胶接点焊接头的A扫描信号和C扫描图像特征,并进行了拉伸-剪切试验。结果表明:通过观察C扫描图像的特征与A扫描信号的变化,能够划分胶接点焊接头的胶层区、热影响区、熔合区、熔核区以及焊接缺陷;随着电流(7~10 k A)的逐渐增大,接头熔核直径呈递增趋势,相应的失效载荷从7 231.5增加到10 939.0 N;当电流为7 k A时,在C扫描图像上反映出飞溅缺陷,此时接头失效载荷远小于没有出现飞溅的接头,可见飞溅降低了胶接点焊接头的拉剪载荷。     

Asymptotic Performance of Sparse Signal Detection Using Convex Programming Method

The detection of sparse signals against background noise is considered. Detecting signals of such kind is difficult since only a small portion of the signal carries information. Prior knowledge is usually assumed to ease detection. In this paper, we consider the general unknown and arbitrary sparse signal detection problem when no prior knowledge is available. Under a Neyman-Pearson hypothesis-testing framework, a new detection scheme is proposed by combining a generalized likelihood ratio test (GLRT)-like test statistic and convex programming methods which directly exploit sparsity in an underdetermined system of linear equations. We characterize large sample behavior of the proposed method by analyzing its asymptotic performance. Specifically, we give the condition for the Chernoff-consistent detection which shows that the proposed method is very sensitive to the 2 norm energy of the sparse signals. Both the false alarm rate and the miss rate tend to zero at vanishing signal-to-noise ratio (SNR), as long as the signal energy grows at least logarithmically with the problem dimension. Next we give a large deviation analysis to characterize the error exponent for the Neyman-Pearson detection. We derive the oracle error exponent assuming signal knowledge. Then we explicitly derive the error exponent of the proposed scheme and compare it with the oracle exponent. We complement our study with numerical experiments, showing that the proposed method performs in the vicinity of the likelihood ratio test (LRT) method in the finite sample scenario and the error probability degrades exponentially with the number of observations.