基于多数据融合传感器的分布式温度控制系统中英文对照.doc

DISTRIBUTED TEMPERATURE CONTROL SYSTEM
BASED ON MULTI-SENSOR DATA FUSION

Cai Yan, Yang Hailan ,Hua Xueming and Wu Yixiong
Abstract:
Temperature control system has been widely used over the past decades. In this paper, a general architecture of distributed temperature control system is put forward based on multi-sensor data fusion and CAN bus. A new method of multi-sensor data fusion based on parameter estimation is proposed for the distributed temperature control system. The major feature of the system is its generality, which is suitable for many fields of large scale temperature control. Experiment shows that this system possesses higher accuracy, reliability, good real—time characteristic and wide application prospect
Keywords:
Distributed control system; CAN bus; intelligent CAN node; multi-sensor data fusion.

1. Introduction
Distributed temperature control system has been widely used in our daily life and production, including intelligent building, greenhouse, constant temperature workshop, large and medium granary, depot, and so on[1]. This kind of system should ensure that the environment temperature can be kept between two predefined limits. In the conventional temperature measurement systems we build a network through RS-485 Bus using a single-chip metering system based on temperature sensors. With the aid of the network, we can carry out centralized monitoring and controlling. However, when the monitoring area is much more widespread and transmission distance becomes farther, the disadvantages of RS-485 Bus become more obvious. In this situation, the transmission and response speed becomes lower, the anti-interference ability becomes worse. Therefore, we should seek out a new communication method to solve the problems produced by RS-485 Bus.
During all the communication manners, the industrial control-oriented field bus technology can ensure that we can break through the limitation of traditional point to point communication mode and build up a real distributed control and centralized management system. As a serial communication protocol supporting distributed real-time control, CAN bus has much more merits than RS-485 Bus, such as better error correction ability, better real-time ability, lower cost and so on. Presently, it has been extensively used in the implementation of distributed measurement and control domains.
With the development of sensory technology, more and more systems begin to adopt multi-sensor data fusion technology to improve their performances. Multi-sensor data fusion is a kind of paradigm for integrating the data from multiple sources to synthesize the new information so that the whole is greater than the sum of its parts [3][4][5]. And it is a critical task both in the contemporary and future systems which have distributed networks of low-cost, resource-constrained sensors
2. Distributed architecture of the temperature control system
The distributed architecture of the temperature control system is depicted in the Figure 1. As can be seen, the system consists of two modules—several intelligent CAN nodes and a main controller. They are interconnected with each other through CAN bus. Each module performs its part into the distributed architecture. The following is a brief description of each module in the architecture.

3．1main controller
As the system’s main controller, the host PC can communicate with the intelligent CAN nodes. It is devoted to supervise and control the whole system, such as system configuration, displaying running condition, parameter initialization and harmonizing the relationships between each part. What’s more, we can print or store the system’s history temperature data, which is very useful for the analysis of the system performance
3.2. Intelligent CAN node
Each intelligent CAN node of the temperature control system includes five units: MCU—a single chip, A/D conversion unit, temperature monitoring unit—sensor group, digital display unit and actuators—a cooling unit and a heating unit. The operating principle of the intelligent CAN node is described as follows.
In the practical application, we divide the region of the control objective into many cells, and lay the intelligent CAN nodes in some of the typical cells. In each node, MCU collects temperature data from the temperature measurement sensor groups with the aid of the A/D conversion unit. Simultaneously, it performs basic data fusion algorithms to obtain a fusion value which is more close to the real one. And the digital display unit displays the fusing result of the node timely, so we can understand the environment temperature in every control cell separately.
By comparing the fusion value with the set one by the main controller, the intelligent CAN node can implement the degenerative feedback control of each cell through enabling the corresponding heating or cooling devices. If the fusion result is bigger than the set value in the special intelligent CAN node, the cooling unit will begin to work. On the contrary, if the fusion result is less than the set value in the node the heating unit will begin to work. By this means we can not only monitor the environment temperature, but also can make the corresponding actuator work so as to regulate the temperature automatically. At the same time every CAN node is able to send data frame to the CAN bus which will notify the main controller the temperature value in the cell so that controller can conveniently make decisions to modify the parameter or not. Since the CAN nodes can regulate the temperature of the cell where they are, the temperature in the whole room will be kept homogeneous. What’s more, we can also control the intelligent node by modifying the temperature’s setting value on the host PC.
Generally, the processors on the spot are not good at complex data processing and data fusing, so it becomes very critical how to choose a suitable data fusion algorithm for the system. In the posterior section, we will introduce a data fusion method which is suitable for the intelligent CAN nodes。

4. Multi-sensor data fusion
The aim to use data fusion in the distributed temperature control system is to eliminate the uncertainty, gain a more precise and reliable value than the arithmetical mean of the measured data from finite sensors. Furthermore, when some of the sensors become invalid in the temperature sensor groups, the intelligent CAN node can still obtain the accurate temperature value by fusing the information from the other valid sensors.
4.1. Consistency verification of the measured data
During the process of temperature measurement in our designed distributed temperature control system, measurement error comes into being inevitably because of the influence of the paroxysmal disturb or the equipment fault. So we should eliminate the careless mistake before data fusion.
We can eliminate the measurement errors by using scatter diagram method in the system equipped with little amount of sensors. Parameters to represent the data distribution structure include median—TM, upper quartile number—Fv, lower quartile number—FL and quartile dispersion—dF.
It is supposed that each sensor in the temperature control system proceeds temperature measurement independently. In the system, there are eight sensors in each temperature sensor group of the intelligent CAN node. So we can obtain eight temperature values in each CAN node at the same time. We arrange the collected temperature data in a sequence from small to large:
T1, T2, …, T8
In the sequence, T1 is the limit inferior and T8 is the limit superior.
We define the median—TM as:
(1)
The upper quartile—Fv is the median of the interval [TM, T8].The lower quartile number—FL is the median of the interval [T1, TM].The dispersion of the quartile is:
（2）

We suppose that the data is an aberration one if the distance from the median is greater than adF, that is, the estimation interval of invalid data is:
(3)
In the formula, a is a constant, which is dependent on the system measurement error, commonly its value is to be 0.5, 1.0, 2.0 and so on.
The rest values in the measurement column are considered as to be the valid ones with consistency. And the Single-Chip in the intelligent CAN node will fuse the consistent measurement value to obtain a fusion result

5. Temperature measurement data fusion experiment
By applying the distributed temperature control system to a greenhouse, we obtain an array of eight temperature values from eight sensors as follows

The mean value of the eight measurement temperature result is

Comparing the mean value (8)T with the true temperature value in the cell of the greenhouse, we can know that the measurement error is +0.5℃. After we eliminate the careless error from the fifth sensor using the method introduced before, we can obtain the mean value of the rest seven data (7)T=29.6℃ the measurement error is -0.4℃.
The seven rest consistent sensor can be divided into two groups with sensor S1, S3, S7 in the first group and sensor S2, S4, S6, S8 in the second one. The arithmetical mean of the two groups of measured data and the standard deviation are as follows respectively

According to formula (13), we can educe the temperature fusion value with the seven measured temperature value.
The error of the fusion temperature result is -0.3℃.
It is obvious that the measurement result from data fusion is more close to the true value than that from arithmetical mean. In the practical application, the measured temperature value may be very dispersive as the monitoring area becomes bigger, data fusion will improve the measuring precision much more obviously.

6. Conclusions
The distributed temperature control system based on multi-sensor data fusion is constructed through CAN bus. It takes full advantage of the characteristics of field bus control system---FDCS. Data acquisition, data fusion and system controlling is carried out in the intelligent CAN node, and system management is implemented in the main controller (host PC). By using CAN bus and data fusion technology the reliability and real-time ability of the system is greatly improved. We are sure that it will be widely used in the future.

References
[1]D.SuandW.McFarland,“A2.5-V,1-WmonolithicCMOSRFpower
amplifier,”inProc.IEEECustomIntegratedCircuitsConf.,1997,pp.
189–192.
[2]D.K.SuandWJ.McFarland,“AnICforlinearizingRFpoweramplifiers
usingenvelopeeliminationandrestoration,”IEEEJ.ofSolid-State
Circuits,vol.33,no.12,pp.2252–2258,Dec.1998.
[3]B.Baliweber,R.Gupta,andD.J.Allstot,“Fully-integratedCMOS
RFamplifier,”inProc.IEEEInt.Solid-StateCircuitsConf.,1999,pp.
72–73.
[4]A.Giry,J.-M.Fourniert,andM.Pons,“A1.9GHzlowvoltageCMOS
poweramplifierformediumpowerRFapplications,”inProc.IEEE
RadioFrequencyIntegratedCircuits(RFIC)Symp.,2000,pp.121–124.
[5]C.FallesenandP.Asbeck,“A1W0.35 mCMOSpoweramplifier
forGSM-1800with45%PAE,”inProc.IEEEInt.Solid-StateCircuits
Conf.,2001,pp.158–159.
[6]R.Gupta,B.M.Ballweber,andD.J.Allstot,“Designandoptimization
ofCMOSRFpoweramplifiers,”IEEEJ.Solid-StateCircuits,vol.36,
no.2,pp.166–175,Feb.2001.

基于多数据融合传感器的分布式温度控制系统
蔡艳，杨海澜，许轲，吴毅雄
摘要：
在过去的几十年，温度控制系统已经被广泛的应用。对于温度控制提出了一种基于多传感器数据融合和CAN总线控制的一般结构。一种新方法是基于多传感器数据融合估计算法参数分布式温控系统。该系统的重要特点是其共性，其适用于很多具体领域的大型的温度控制。实验结果表明该系统具有较高的准确性、可靠性，良好的实时性和广泛的应用前景。
关键词：
分布式控制系统；CAN总线控制；智能CAN节点；多数据融合传感器。
1介绍
分布式温度控制系统已经被广泛的应用在我们日常生活和生产,包括智能建筑、温室、恒温车间、大中型粮仓、仓库等。这种控制保证环境温度能被保持在两个预先设定的温度间。在传统的温度测量系统中，我们用一个基于温度传感器的单片机系统建立一个RS-485局域网控制器网络。借助网络,我们能实行集中监控和控制.然而,当监测区域分布更广泛和传输距离更远,RS-485总线控制系统的劣势更加突出。在这种情况下，传输和响应速度变得更低，抗干扰能力更差。因此，我们应当寻找新的通信的方法来解决用RS-485总线控制系统而产生的问题。在所有的通讯方式中，适用于工业控制系统的总线控制技术，我们可以突破传统点对点通信方式的限制、建立一个真正的分布式控制与集中管理系统，CAN总线控制比RS-485总线控制系统更有优势。比如更好的纠错能力、改善实时的能力，低成本等。目前，它正被广泛的应用于实现分布式测量和范围控制。
随着传感器技术的发展，越来越多的系统开始采用多传感器数据融合技术来提高他们的实现效果。多传感器数据融合是一种范式对多种来源整合数据,以综合成新的信息，比其他部分的总和更加强大。无论在当代和未来，系统的低成本，节省资源都是传感器中的一项重要指标。
2分布式架构的温度控制系统
分布式架构温度控制系统如图中所示的图1。可以看出，这系统由两个模块——两个智能CAN节点和一个主要的控制器组成。每个模块部分执行进入分布式架构。下面的是简短的描述下各模块。
3.1主要控制器
作为系统的主要控制器，这主pc能和智能CAN节点通信。它致力于监督和控制整个系统，系统配置、显示运行状况、参数初始化和协调各部分间的关系。更重要的是，我们能打印或储存系统的历史温度的数据，这对分析系统性能是非常有用的。
3.2智能CAN节点
每一个温度控制系统的智能CAN节点有五个部分：MCU—一个单片机，A/D转换单元，温度监测单元—传感器群，数字显示器，激发器—一个冷却单元和供暖单元。接下来介绍智能CAN节点的工作原理。
在实际操作中，我们划分控制的目标进入一些单元，储存智能CAN节点在一些典型的单元。在每个节点，单片机借助A / D转换单位从温度测量传感器收集温度数据。同时，它执行基本的数据融合运算获得运算的结果,更接近实际。数字显示器及时显示融合节点的结果，所以我们能及时了解在每个控制单元所处的环境温度。
通过比较融合值用主控制器构建一个，这样智能CAN节点可以通过相应的加热或冷却装置实现反馈控制各单元。如果在特别的智能CAN节点融合结果大于设定值，冷却单位将开始工作。相反,如果在节点融合的结果低于设定值加热单位将开始工作。用这种方法,我们不仅能监控环境温度,还能做相应的触发器来实现温度的自动调节。与此同时，每个CAN节点发送数据帧到CAN总线，CAN总线将告知在着单元中的主控制器这温度值，那么这控制器能便利的作出是否修改这参数的决定。自从这CAN节点有调节温度的单元在，整个房间的温度将保持均匀。更重要的是,我们也可以通过在主pc上修改温度的设定值来控制这智能节点。
一般来说,处理器不擅长即时的复杂的数据处理和数据融合,所以如何选择合适的数据融合算法对系统变得至关重要。后一节中,我们将介绍适合于智能CAN节点的数据融合方法。
4.多传感器数据融合
旨在利用数据融合在分布式温度控制系统中来消除不确定性,获得更精确、可靠是比从限定的传感器的测量数据的算数平均值更重要。当一些传感器的温度传感器变为无效的，这智能CAN节点还可以通过熔断这些信息而从有用的传感器获得精确温度。
4.1实测数据的一致性核实
在我们设计的分布式温度控制系统的温度测量的过程中,突发性干扰或设备故障的影响不可避免的产生测量误差。所以在数据融合前我们应该消除错误的误差。
我们可以利用系统中配备的少量传感器用散点图发消除这个测量误差。用参数来代表数据分布结构包括中值——TM,上四位数—— Fv,下四位数——FL和分散四位数——dF.
人们认为每个传感器在温度控制系统的温度测量所得独立。在系统中，有八个传感器在各智能CAN节点的温度传感器群。所以我们在每个CAN节点同一时刻能获得8个温度值。我们安排收集到的温度数据序列由小到大:
T1, T2, …, T8
在序列中，T1是最低位而T8是最高位。我们定义TM为：

上四位数——Fv是区间[TM, T8]的中值，低四位数—— Fl是区间[T1, TM]的中值，这四位数的离散是：。
该公式,一个是常数,取决于系统测量误差, 通常值是0.5,1.0,2.0等等。在数列中其余的测量值都被看作是于有效值一致的。在智能CAN节点的单片机智将把一致的测量值融合。
5. 温度测量的数据融合的举例
分布式温度控制系统运用于一间温室, 我们从8个温度传感器获得一组8个温度值如下

八个温度测量值的结果

把在这温室中的八个温度的平均值和真实的温度值做比较，我们可以知道测量误差是+ 0.5℃。之后在介绍这个方法前我们消除从这第五个传感器的测量误差，我们能得到的剩余的七个数据的平均值(7)T = 29.6℃, 测量误差是-0.4℃.这剩下的七个传感器被分成两个传感器组，S1, S3, S7 是第一组，S2, S4, S6, S8 是第二组。两组测量数据的算术平均和标准偏差分别如下：

根据公式(13), 我们可以用七个测量温度确定温度融合值。
融合温度的结果的误差是-0.3℃。
很明显,数据融合测量结果比算术的平均值更接近于实际值。在实际操作中,测量温度可能是很分散的变得更大的监测区域,数据融合将更加明显提高了测量精度。
6.总结
这基于多数据融合传感器的分布式温度控制系统是通过CAN总线构建。它充分利用了FDCS即时总线控制系统的特点。数据采集，数据融合，系统控制用智能CAN节点得到实现，而系统管理通过主控制器（host PC）被实现。通过使用CAN总线与数据融合技术系统的可靠性和实时的能力被大大提高了。我们确定它在将来会得到广泛的应用。
目 录
第一章 总论 5
1.1项目概要 5
1.2可行性研究报告编制依据 8
1.3综合评价 8
第二章 项目背景及必要性 11
2.1项目建设背景 11
2.2项目建设的必要性 13
第三章 建设条件 16
3.1项目区概况 16
3.2项目建设条件优劣势分析 21
第四章 市场分析与销售方案 26
4.1市场分析 26
4.2销售策略、营销方案和模式 29
4.3风险分析 30
第五章 项目建设方案 32
5.1建设任务和规模 32
5.2建设规划和布局 32
5.3工艺（技术）方案 32
5.4建设内容 35
5.5实施进度安排 36
第六章 环境影响评价 38
6.1环境影响 38
6.2环境保护与治理措施 38
第七章 项目组织与管理 40
7.1组织机构与职能划分 40
7.2劳动定员 40
7.3经营管理模式 41
7.4经营管理措施 41
7.5技术培训 42
7.6劳动保护、安全卫生与消防 42
第八章 投资概算与资金来源 44
8.1投资概算依据 44
8.2投资概算 44
8.3资金来源 46
第九章 财务评价 47
9.1财务评价依据 47
9.2销售收入、销售税金和附加估算 47
9.3总成本及经营成本估算 48
9.4财务效益分析 49
9.5不确定性分析 50
9.6财务评价结论 51
第十章 社会效益评价 52
10.1社会评价基本结论 52
10.2农业产业化经营 52
10.3农民增收、农业增效评价 54
10.4其它社会影响 55
第十一章　可行性研究结论与建议 56
11.1可行性研究结论 56
11.2问题与建议 56